ACS Publications. Most Trusted. Most Cited. Most Read
Systematic Comparison of Experimental Crystallographic Geometries and Gas-Phase Computed Conformers for Torsion Preferences
More

OR SEARCH CITATIONS

My Activity
Publications

Figure 1Loading Img
Download Hi-Res ImageDownload to MS-PowerPointCite This:J. Chem. Inf. Model. 2023, 63, 23, 7401-7411
  • Open Access
Computational Chemistry

Systematic Comparison of Experimental Crystallographic Geometries and Gas-Phase Computed Conformers for Torsion Preferences
Click to copy article linkArticle link copied!

  • Dakota L. Folmsbee
    Dakota L. Folmsbee
    Department of Chemistry, University of Pittsburgh, 219 Parkman Avenue, Pittsburgh, Pennsylvania 15260, United States
    Department of Anesthesiology & Perioperative Medicine, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, United States
    More by Dakota L. Folmsbee
  • David R. Koes
    David R. Koes
    Department of Computational & Systems Biology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, United States
    More by David R. Koes
    Orcidhttps://orcid.org/0000-0002-6892-6614
  • Geoffrey R. Hutchison*
    Geoffrey R. Hutchison
    Department of Chemistry, University of Pittsburgh, 219 Parkman Avenue, Pittsburgh, Pennsylvania 15260, United States
    Department of Chemical & Petroleum Engineering, University of Pittsburgh, 3700 O’Hara Street, Pittsburgh, Pennsylvania 15261, United States
    *Email: [email protected]
    More by Geoffrey R. Hutchison
    Orcidhttps://orcid.org/0000-0002-1757-1980
Open PDFSupporting Information (1)

Journal of Chemical Information and Modeling

Cite this: J. Chem. Inf. Model. 2023, 63, 23, 7401–7411
Click to copy citationCitation copied!
https://doi.org/10.1021/acs.jcim.3c01278
Published November 24, 2023

Copyright © 2023 The Authors. Published by American Chemical Society. This publication is licensed under

CC-BY 4.0 .

Abstract

Click to copy section linkSection link copied!
High Resolution Image
Download MS PowerPoint Slide

We performed exhaustive torsion sampling on more than 3 million compounds using the GFN2-xTB method and performed a comparison of experimental crystallographic and gas-phase conformers. Many conformer sampling methods derive torsional angle distributions from experimental crystallographic data, limiting the torsion preferences to molecules that must be stable, synthetically accessible, and able to be crystallized. In this work, we evaluate the differences in torsional preferences of experimental crystallographic geometries and gas-phase computed conformers from a broad selection of compounds to determine whether torsional angle distributions obtained from semiempirical methods are suitable priors for conformer sampling. We find that differences in torsion preferences can be mostly attributed to a lack of available experimental crystallographic data with small deviations derived from gas-phase geometry differences. GFN2 demonstrates the ability to provide accurate and reliable torsional preferences that can provide a basis for new methods free from the limitations of experimental data collection. We provide Gaussian-based fits and sampling distributions suitable for torsion sampling and propose an alternative to the widely used “experimental torsion and knowledge distance geometry” (ETKDG) method using quantum torsion-derived distance geometry (QTDG) methods.

This publication is licensed under

CC-BY 4.0 .
  • cc licence
  • by licence
Copyright © 2023 The Authors. Published by American Chemical Society

Subjects

what are subjects

Introduction

Click to copy section linkSection link copied!
Most molecules exhibit some level of conformational flexibility, the existence of multiple low-energy geometries that differ mostly by changes in the torsional angles of both acyclic and ring bonds. Many methods have been developed to sample conformations, with benchmarks frequently focusing on finding one geometry close to an experimental crystal structure. (1−4) Consequently, most conformer sampling methods derive torsional angle distributions from experimental crystallographic data (1,5−8) not only to provide geometries close to such benchmarks but also as large diverse repositories of “ground truth” geometric properties such as bond lengths, angles, and dihedrals. (9−11)
One challenge is that experimental crystallographic data are limited by the size of the data source (12) and reflect some inherent biases. In order to be collected, the molecules must be stable, synthetically accessible, and actually made and crystallized. While new cryo-electron microscopy (cryo-EM) techniques are improving dramatically and have less stringent requirements on crystals, generally, growing high-quality crystals for small-molecule crystallography is a time-consuming process. Moreover, it is known that compounds with experimental crystal structures are generally smaller and exhibit fewer conformers than other compounds. (13) Similarly, compounds containing elements outside the common organic subset (e.g., B, As, and Se) or less common chemical motifs may be poorly represented in experimental crystallographic databases. Also, even for compounds found in an existing database, much chemistry is performed in solution and gas phases, where solid-state preferences may not directly apply. (14−18) Finally, several works have noted challenges with deriving data from some crystallographic databases. (12,19)
Consequently, finding unbiased alternative sources of accurate and reliable torsional angle preferences could significantly expand the use of conformational sampling to a new chemical space. Typically, sampling has been performed using small-molecule force fields (e.g., UFF (20)), which have shown limited fidelity compared to density functional and other first-principles quantum chemical methods. (21,22) The development of efficient dispersion-corrected semiempirical methods such as GFN2-xTB, (23) as well as new machine learning methods such as ANI (24−27) and OrbNet, (28,29) offers improved accuracy of torsional angles and nonbonded interactions with moderate computational cost. Moreover, several large-scale computational efforts including PubChemQC (30) and the QCArchive (31,32) have provided large amounts of optimized gas-phase theoretical geometries using high-quality density functional methods.
In this work, we outline an extensive effort to analyze the conformers and torsional angle preferences of more than 3 million organic small molecules, using exhaustive sampling using the GFN2-xTB method across both the experimental Crystallographic Open Database (33) (COD) and multiple sets of small molecules, including PubChemQC. (30) We compare the potential bias between the crystallographic and gas-phase geometries and individual torsion patterns, including analysis with ωB97X-D3 (34) with the def2-SVP basis set. (35,36)

Methods

Click to copy section linkSection link copied!
Molecules for this work were compiled from several sources, including 88,106 organic compounds from the Crystallographic Open Database, (9,10) 3,009,591 molecules from PubChemQC, (30) 88,550 molecules from the Pitt Quantum Repository, and previous work on conformational flexibility, which included 70,850 molecules from a subset of ZINC (37) and 4378 molecules from the Platinum ligand database. (3) For all sources, the largest substructure was retained (i.e., the solvent or salts were removed from the crystallographic unit cells). For compounds without initial 3D coordinates, Open Babel 3.1 (38−40) was used to generate initial coordinates, since CREST requires an initial geometry. As noted above, the total set of compounds included more than 3 million unique molecules. Properties across the data sets are illustrated in Figures S1–S4 and are roughly comparable to the Platinum Diverse set and the DrugBank-approved set, (41) with the exception that the PubChemQC set compounds are smaller.
For each molecule, conformers were generated using the CREST program to exhaustively sample the potential energy surface, using default parameters and the GFN2-xTB method to compute energies (hereafter known as simply GFN2) and optimize geometries. (23,42,43) In some cases, CREST produced fragments or chemical rearrangements (e.g., producing different compounds than the input, based on the InChI identifier)─these systems were excluded from analysis. While, in principle, CREST can generate many conformers per compound, we find that for the vast majority of compounds, only a few conformers are generated within 6 kcal/mol as calculated by GFN2 (Figures S5 and S6), consistent with our previous work. (44)
In this work, the lowest-energy conformer by GFN2 energy was analyzed. Torsional angle SMILES arbitrary target specification (SMARTS) patterns from the experimental torsion knowledge-based distance geometry (ETKDG) approach (6−8) were used to generate the histograms for the gas-phase data. ETKDG derives torsional preference distributions for the RDKit distance geometry coordinate generation method from the analysis of experimental crystallographic data for a set of hierarchical dihedral patterns. (6−8)
These patterns are constructed from molecules with central bonds of C–C (168 acyclic patterns), C–O (56 acyclic patterns), C–S (16 acyclic patterns), N–C (131 acyclic patterns), N–S (4 acyclic patterns), and S–S (1 pattern) and create the 387 acyclic and 105 ring dihedral SMARTS patterns used to generate histograms of matching torsions with a stepsize of 5° using RDKit (45) Python scripts (see the Supporting Information). Figures depicting the SMARTS patterns were generated using SMARTS.plus. (46,47)
For selected compounds, to compare the GFN2 geometry with density functional theory (DFT), optimization was performed with ORCA 4.2.0 (48) using the ωB97X-D functional (34) and the def2-SVP basis set, (35,36) which has proven to produce fairly accurate conformational energetics (21,49,50) although some errors still exist when compared with more accurate methods. (51−54)

Results and Discussion

Click to copy section linkSection link copied!
Traditionally, conformer sampling is refined via classical force fields, which yield a poor correlation with energies from more accurate quantum methods. (21,22) On the other hand, geometry optimization with most density functional methods requires hours per conformer, making large-scale sampling prohibitive. Recently, larger data sets have emerged, including this work, as well as the ANI-1x, (55) GEOM, (56) QCArchive, (31,32) QMugs, (57) and SPICE sets. (58) The goal of this work is to show that while gas-phase quantum chemical geometries require substantial time to generate, such data sets can be used to augment or supplant traditional crystal structure sources to establish torsional preferences.
Below, we will consider the overall molecular geometries between optimized gas-phase conformer ensembles and experimental crystal structures, individual torsional distributions, compare the GFN2-optimized and DFT-optimized geometries, and finally fit the torsional distributions via Fourier analysis or sets of Gaussian peaks.

Comparing Overall Geometries

Often, coordinate generation and conformer tools are evaluated, in part, by comparing the ensemble of generated geometries to experimental crystallographic geometries. For example, the ETKDG method has proven to generate structures with small root-mean-square displacement (RMSD) when compared to small-molecule crystal structures and bound-ligand geometries. (1,3,6,7)
Consequently, our first comparison is between the CREST-generated GFN2 conformer ensembles and ETKDG 250 conformer ensembles across the Crystallographic Open Database (COD) and Platinum Diverse data sets. As noted above, while CREST attempts to generate exhaustive ensembles under 6 kcal/mol from the global minima, in general, only a few conformers are generated (Figures S5 and S6), and 250 conformers is sufficient to encompass 93% of the COD set and 85% of the Platinum Diverse set (Figure S5). While larger ensembles tend to yield smaller RMSD, a comparison is still useful despite some differences in ensemble size. (59)
Calculating the smallest non-hydrogen RMSD of each generated ensemble and the experimental crystal structure geometry (Figure 1), we find that the CREST ensembles perform better than ETKDG on molecules with few rotatable bonds (e.g., from zero to ∼3–4 rotors) on the COD set, have comparable RMSD across a broader range of COD molecules, and perform slightly worse on the Platinum set (e.g., Figure 1a,b and Table S1). Note that the closest geometry to the experimental geometry is often not the lowest-energy CREST/GFN2 conformer.

Figure 1

Figure 1. Comparison of the smallest non-hydrogen RMSD between experimental crystallographic geometry and CREST or ETKDG conformers for the (left) Crystallographic Open Database (COD) and (right) Platinum Diverse data set. (c–f) Captions indicate best-fit linear regression in Å. Note that for both data sets, CREST produces smaller RMSD for molecules with few rotatable bonds but a larger slope indicating generally worse RMSD for larger compounds with more rotatable bonds.

High Resolution Image
Download MS PowerPoint Slide
As illustrated in Figure 1 and summarized in Tables S1 and S2, for the COD small-molecule crystal structures, the CREST/GFN2 method has a higher fraction of molecules within an RMSD of 0.2–0.5 Å, compared to ETKDG, and a smaller median RMSD. Such performance arises mostly from molecules up to 3–4 rotatable bonds. The improved treatment of nonbonded interactions and electrostatics in the quantum GFN2 method likely gives rise to these geometries more closely matching experimental crystallographic geometries. (21)
Note, however, that for the Platinum set, derived from bound PDB ligands, the performance of the CREST ensembles is generally worse than that of ETKDG, with a worse median and mean RMSD and a considerably higher RMSD increase with the number of rotatable bonds, as illustrated in Figure 1. We speculate that this difference derives from comparing gas-phase CREST/GFN2 conformers with bound-ligand geometries, which may be more stabilized in extended conformations due to intermolecular interactions with a binding site and solvent. (60)
While there is a wide distribution of the calculated radius of gyration between the lowest-energy CREST/GFN2 generated conformations and the experimental crystal structures from COD and Platinum sets (Figure 2), as compiled in Figure 2, the median radii of gyration for experimental COD and CREST-generated conformers are relatively close (e.g., the ratio of the two is usually close to 1.0, Figure S7). On the other hand, the Platinum crystallographic geometries show a notably larger radius of gyration than the CREST conformer (e.g., Figure S7), with the deviation growing as a function of the number of rotatable bonds.

Figure 2

Figure 2. Calculated radius of gyration for the lowest-energy CREST conformer and experimental geometries as a function of the number of rotatable bonds for the (a) Crystallographic Open Database (COD) and (b) Platinum data sets and scatterplot of compounds from (c) COD and (d) Platinum sets, comparing the radius of gyration from the CREST/GFN2 lowest-energy conformation with that of the experimental crystal structure geometry. Dashed line indicates a 1:1 correspondence, with approximate bounds indicated by solid lines.

High Resolution Image
Download MS PowerPoint Slide
Overall, the results indicate that CREST-derived conformer ensembles perform comparably to ETKDG ensembles on the small-molecule COD set, with the caveat of a potential bias toward compact conformations relative to the Platinum bound PDB ligands, likely due to a neglect of intermolecular interactions in the gas-phase geometries.
While CREST gas-phase ensembles perform acceptably compared to ETKDG on the small-molecule COD set, the time required for the calculations is large, as illustrated in Figure S8. The median runtime across the COD set is 1 h and increases as n2 (62) with n as the number of atoms as both the semiempirical GFN2 geometry optimizations increase in time and larger molecules generally yield more conformers (e.g., Figure S6). Consequently, the approach discussed below is to use a database of CREST ensembles to derive torsional preferences suitable for developing faster conformational search tools.

Individual Torsion Preferences

As mentioned above, traditional efforts to derive torsional preferences use experimental crystallographic databases. (1,3,5−8) Given the reasonable agreement between CREST/GFN2 conformers and crystallographic geometries, particularly on the small-molecule COD set, the individual torsion preferences should also be comparable. Thus, we compared the experimental torsions from the COD to the lowest-energy gas-phase CREST/GFN2 generated conformers from the COD, as well as to CREST/GFN2 generated conformers of all combined sets (e.g., PubChemQC, COD, etc.) across over 3 million compounds.
Comparing torsions between experimental geometries from the COD and lowest-energy gas-phase conformers from CREST/GFN2 is intended to show the suitability of gas-phase torsion preferences to replace or supplement standard crystallographic analysis. Poor correlation between individual torsion distributions could occur for a few main reasons─that the gas phase and experimental crystal structures show distinct angles due to intermolecular interactions and packing effects, that the distributions have few points and are thus inherently noisy, or that the GFN2 semiempirical method is not sufficiently accurate to reproduce torsion angles. (51,61)
However, the COD is significantly smaller than the Cambridge Crystallographic Database, (11,62) which limits the number of torsion data points for some structural motifs. Expanding our analysis to include millions of molecules can improve the understanding of uncommon motifs as well as potential torsional preference differences in the solid state compared to the gas-phase geometries. This work will examine the comparison of crystal structure and calculated gas-phase torsional preferences for both ring- and acyclic-containing torsions.
To determine the degree of correlation between the COD experimental torsion preferences and the computed gas-phase preferences, the r2 of the kernel density estimation (KDE) for both acyclic and ring torsion patterns was compiled, as shown in Figure 3a. The KDE is used to smooth the histogram (i.e., representing uncertainty in the individual torsions) and to avoid issues when correlating areas with few or no torsions in the COD with the regions where some torsions were present in the combined set. The r2 then correlates the KDE of the experimental COD torsion preferences and the KDE of the computed CREST/GFN2 gas-phase preferences for each torsion pattern. A total of 127 patterns yield r2 > 0.8 and the median acyclic r2 is 0.61, indicating that while there are some differences, the torsional preferences yield a reasonable correlation.

Figure 3

Figure 3. Correlation between experimental and gas-phase torsions across the COD data set for (a) acyclic patterns and (b) ring patterns.

High Resolution Image
Download MS PowerPoint Slide
The correlation is likely better indicated by the median value, as shown by analyzing the 76 patterns with an r2 less than 0.2. The patterns in this regime had a median of only 175 instances across the COD experimental set, which appears to be too few to form accurate distributions of the dihedral angle preferences.
In short, we believe that in most cases, the correlation between experimental crystallographic geometries and gas-phase conformer sampling with semiempirical quantum methods such as GFN2 is high enough to derive accurate torsional preferences. This is consistent with previous comparisons of energetic rankings of conformers between GFN2 and higher-level quantum chemical methods including density functional and coupled cluster calculations, (21) as well as efforts to estimate torsional strain in crystal structures using DFT methods. (61,63,64)
Ring torsional patterns, constrained by the nature of a ring, show even greater correlation, with the median r2 at 0.83 (Figure 3b) and few patterns (16 out of 105) showing correlations below 0.5, again correlated with few matches.
Individual acyclic torsional preferences can be further analyzed to determine the qualitative correlation between the crystal structures and gas-phase conformers. Torsion pattern 229 shown in Figure 4 demonstrates a high degree of correlation (r2 = 0.76) between the crystal structures and gas-phase CREST/GFN2 lowest-energy conformers while demonstrating the advantage of using additional data from the entire set of compounds. Pattern 229 has only 179 torsions from the COD, while the expanded data set boasts over 25,000 instances. This increase in data clarifies torsional preferences in the range of 90–150° as the conformers demonstrate a clearer peak in this region and less noise overall.

Figure 4

Figure 4. Histograms for pattern 229, including COD experimental torsional angles, gas-phase lowest-energy conformers from the same COD molecules, and gas-phase lowest-energy conformers across the entire data set, indicating the strong correlations and that the increased quantity of data greatly refines the histograms.

High Resolution Image
Download MS PowerPoint Slide
Figure 5 demonstrates how the lack of data in the COD can impact qualitative assessments of the torsional preferences. The correlation between the experimental COD and computed gas-phase torsions (left and middle panels) is 0.58. The experimental torsions show a mild preference around 60 and 120°, with a lot of noise between, while the calculated torsions have more defined peaks at 50, 90, and 132°. Much like the case above, the 10× increase in the number of torsions in the combined set was able to provide enough data to discern preferential angles.

Figure 5

Figure 5. Histograms for pattern 307, including COD experimental torsional angles, gas-phase lowest-energy conformers from the same COD molecules, and gas-phase lowest-energy conformers across the entire data set, again indicating how the increased quantity of data greatly refines the torsional preferences (e.g., peaks near 50, 90, and 132°).

High Resolution Image
Download MS PowerPoint Slide
In addition to the acyclic preferences, we analyzed the preferences of ring torsional patterns. Ideally, torsional patterns in rings should correlate well between experimental and calculated gas-phase geometries due to the steric constraints on the geometry. The correlation was analyzed in the same manner as above by taking the r2 of the kernel density estimation (KDE) for each ring torsion pattern and compiling them into Figure 3b. Compared with the median r2 of 0.61 for the acyclic patterns, the ring patterns boasted a median r2 of 0.83, indicating a significant correlation between the two sets.
Similar to the case for acyclic patterns, the increase in available data bolsters the angular preferences of the ring torsion patterns. Figure 6 exemplifies this by demonstrating how the increase in the number of torsions for ring torsion pattern 50 affects the torsional preference. The correlation between the experimental and gas-phase COD data is 0.79. Although the experimental data suggests 0, 60 and 180° are prominent angles; the additional data from the combined sets firmly demonstrates that these are prominent angles for the torsion pattern, with a small subpopulation at ∼41°. The addition of more data exhibits a more complete picture of torsional preferences that better represent the chemical space.

Figure 6

Figure 6. Histograms for COD experimental torsional angles, gas-phase lowest-energy conformers from the same COD molecules, and conformers across the entire data set, again indicating that the increased quantity of data greatly refines the torsional preferences (e.g., a strong peak at 60°).

High Resolution Image
Download MS PowerPoint Slide
These expected similarities for the ring patterns are due to the intermolecular constraints that the ring structure imposes on the geometry. There is less flexibility in ring systems. The correlation of individual ring torsions demonstrates the accuracy of the quantum torsional preferences and indicates the ability to use this information as an additional method for determining the torsional preferences of structural motifs not yet examined through experimental means.
Importantly, sampling ring torsion angles in isolation is inherently challenging since such angles are strongly correlated with other torsion angles in the same ring. (65) In the ETKDG method, such a correlation is resolved through the distance geometry methods, refining distances between all atoms in the ring (and molecule) together. As such, the ring torsion data in this work is intended either for refining such distance geometry methods or as an initial effort to refine ring puckering distributions using Cremer–Pople schemes. (65)
The outlined work demonstrates the desire for additional methods based on quantum calculations where crystal structure constraints may not be suitable or data may not be prevalent due to experimental limitations. Using the example of ETKDG, (7) an alternative, quantum torsion distance geometry (QTDG) could be useful for gas-phase applications. (66) Moreover, a QTDG method would no longer be constrained to structural preferences derived from what can be synthesized and crystallized. This allows for an increase in the capability of the method because a larger amount of data would be available for a more diverse representation of chemical space.

Comparisons between GFN2- and DFT-Optimized Geometries

As noted above, some differences between the COD crystallographic torsional preferences and those calculated from gas-phase CREST/GFN2 lowest-energy conformers derive from patterns with a few candidates and thus somewhat noisy histograms. In other cases, the histograms suggest noticeable differences involving new peaks, particularly those from the combined set. For example, in Figure 7, the experimental torsions show a mix of torsions in the range of 70–110°, while the calculated torsions from the combined set also show a dominant peak at 90°.

Figure 7

Figure 7. Differences in torsional preferences for experimental and gas-phase geometries.

High Resolution Image
Download MS PowerPoint Slide
This new peak could derive from a new subpopulation, for example, compounds not present in the COD or not synthetically accessible or amenable to crystallization. The new peaks could also derive from problems with the GFN2 method, predicting incorrect torsion angles.
To test the latter hypothesis, ωB97X-D/def2-SVP geometry optimizations were performed to verify the GFN2-optimized geometries. Five molecules containing a torsion pattern of 270° and an angle of ∼90° were randomly selected from the combined set. The DFT-optimized geometries were found to be in strong agreement (e.g., within a mean absolute deviation of 1.54°) with the GFN2 torsion angles. Such compounds include steric constraints restricting the torsion to ∼90°, as illustrated by Figure 8.

Figure 8

Figure 8. Example of compound matching torsion pattern 270 with steric constraint forcing an angle of ∼90°. Figure from Avogadro2. (67,68)

High Resolution Image
Download MS PowerPoint Slide
We compared ten random patterns with obvious subpopulations (e.g., 90° in Figure 7), including patterns 10, 23, 48, 53, 125, 149, 216, 253, 270, and 306. For each pattern, five molecules were selected, with the exception of pattern 23, which had only one molecule in the subpopulation. As illustrated in Figure S10, overall absolute torsion angle deviations between the GFN2-optimized and ωB97X-D/def2-SVP-optimized geometries were between 0 and 5°, with the median absolute deviation of 2.89° and a mean absolute deviation of 4.88°.
Comparing across torsion patterns, Table S3 and Figure S11 indicate that some patterns, particularly 48 and 253, reflect larger absolute torsion angle differences of 15 and 10°, respectively, but the overall r2 correlation between GFN2-optimized and ωB97X-D/def2-SVP torsion angles is high, even for these patterns.
Consequently, while GFN2 is an approximate semiempirical method and correlations with more accurate quantum chemical methods are imperfect, (21) we can conclude that differences between the crystallographic and gas-phase torsional preferences appear to be mainly derived from lack of data and the presence of new subpopulations in the combined set.

Fitting and Sampling Torsion Angles

The ETKDG methods derive torsional preferences using a potential energy term fit to a discrete cosine Fourier analysis. (6,7)
V(ϕ)=∑i=16Ki[1+cos(di)cos(iϕ)]
(1)
in which K is a force constant and d is a phase shift and limited to 0 and π arbitrarily. Consequently, we compare the original published ETDKG fits to a similar cosine analysis of our data. In our work, the phase shift is also varied as a parameter by using nonlinear curve fitting. Note that for both the ETKDG and cosine fits, the histogram probabilities must be transformed into relative energies before fitting. We compare the ETKDG and cosine fits to a sum of up to six Gaussian peaks. The code for the nonlinear curve fitting, via SciPy, (69) is included at https://github.com/hutchisonlab/quantum-torsions.
Compiled in Figure 9 are histograms of correlations among the ETKDG, cosine, and Gaussian fits and the underlying torsional distributions for both acyclic and ring patterns. The Pearson r2 correlation is used since the magnitudes of the peak heights may vary, but the key features should be the relative intensities at each torsion angle.

Figure 9

Figure 9. Histograms of correlation between (a,d) ETKDG, (b,e) cosine fits, or (c,f) Gaussian fits and derived torsional histograms for (a–c) acyclic and (d–f) ring torsional patterns.

High Resolution Image
Download MS PowerPoint Slide
Table 1 indicates the median r2 for each fit across both acyclic and ring patterns, comparing the correlation of relative probabilities at each torsion angle between the fit and the underlying histogram. Note that since the cosine fits in this work allow the phase shift parameter to vary, the results are improved over the published ETKDG fits. We speculate that the fixed phases in the previous ETKDG fits are insufficiently physical, and while it may be useful conceptually to consider these distributions as arising from a typical cosine series potential energy function, this assumption does not fit the present data well.
Table 1. Median r2 Correlation between Fit Functions and the Underlying Quantum Torsions for Both Acyclic and Ring Patterns
methodmedian acyclic r2median ring r2
ETKDG0.260.03
cosine fits0.710.73
Gaussian fits0.910.93
Nevertheless, simple fits to Gaussian peaks perform noticeably better, with most fits yielding r2 values above 0.9. In part, this method enables better fits of subpopulations, as discussed above. Further refinement of torsion patterns may resolve such subpopulations, improving the accuracy of ETKDG-style fits. (5)
For sampling torsion angles from the Gaussian fits, we have generated the cumulative sum across all angles from 0 to 360°, normalized and inverted such that generating a uniform random number yields an appropriate torsion angle for a given torsion pattern. Such distributions can either be used for individual torsion driving, as demonstrated in the included script, or for Bayesian sampling, (70,71) or as part of distance geometry coordinate generation methods such as ETKDG.
Finally, we note that while most previous work has compared computed conformers with experimental crystal structures or bound-ligand geometries, recent work has used cross-sectional areas (14,15) and rotational spectroscopy (16−18) to gain geometric insight into gas-phase conformer geometries. Future work should consider such alternate benchmarks since bound conformations, in particular, reflect stabilization from intermolecular interactions. (59)

Conclusions

Click to copy section linkSection link copied!
Methods such as CREST/GFN2 require a median of 1 h per compound, even using a relatively fast semiempirical quantum method. Consequently, its use for widespread conformational sampling will be limited. Deriving data from experimental crystallographic databases, however, requires far more than 1 h per novel compound to synthesize, crystallize, and analyze via X-ray diffraction and is biased by the need for synthesis and crystallization. This work analyzes over 3 M unique compounds, over twice the reported size of the Cambridge Crystallographic Database.
Overall, comparing geometric RMSD between experimental small-molecule crystal structure geometries and gas-phase GFN2 conformer ensembles indicates a high fidelity compared to the widely used ETKDG methods. Particularly for molecules with few rotatable bonds, the CREST/GFN2 ensembles yielded smaller RMSD than ETKDG and comparable performance across larger molecules from the Crystallographic Open Database. We note somewhat worse performance across the Platinum set than that of ETKDG, particularly for molecules with more rotatable bonds. There is a notable difference in the radius of gyration between the Platinum and low-energy CREST/GFN2 compounds, particularly with increasing numbers of rotatable bonds, likely indicating some preference for extended conformations in the Platinum set due to intermolecular interactions of bound ligands, not present in an isolated gas-phase calculation. (59)
Comparing individual torsion preferences between the experimental COD geometries and the lowest-energy CREST/GFN2 conformers indicated good correlation (i.e., median r2 above 0.6 for acyclic torsions and above 0.8 for ring torsions), with most cases of poor correlation derived from few examples in the COD set.
The advantage of gas-phase conformer sampling, particularly with the semiempirical GFN2 method, is that many compounds can be analyzed beyond those available in crystallographic databases. Consequently, distributions of torsion preferences were analyzed across over 3 million compounds compiled from multiple small-molecule sets.
Although some differences in torsional preferences were found, most cases of poor correlation between experimental COD geometries and lowest-energy CREST/GFN2 conformers occur with a lack of sufficient experimental crystallographic data. Some deviations occur from new subpopulations not present in the experimental databases as not every compound can be crystallized. (13)
Using density functional methods to optimize compounds in such subpopulations, we find torsion angle deviations of ∼0–5°, suggesting that CREST/GFN2 is sufficiently accurate for generating overall torsion angle distributions, even if some errors remain.
Finally, fits to Gaussian peaks were used to generate sampling distributions and proved to be more accurate than the Fourier analysis used in ETKDG, likely due to the presence of such subpopulations. Further refinement of torsion patterns will be useful, (5) as well as analysis of correlated torsional preferences for both acyclic and ring torsions. (44,71)
This work has demonstrated the ability of CREST/GFN2 to provide accurate and reliable torsional preferences that could provide a basis for the quantum torsion distance geometry (QTDG) method. Such a method could provide an alternative to ETKDG that does not rely on experimental crystal structure elucidation as well as a method designed particularly for conformer sampling for gas-phase applications rather than targeting crystal structure or bound-ligand geometries that include intermolecular interactions. Finally, QTDG could be used as part of the automated refinement of torsion patterns (5) since additional conformer sampling can be performed to refine patterns with sparse representation in experimental databases, including charged, novel, or hard-to-crystallize species. (13) Future work can also carry out comparisons with geometric information from gas-phase spectroscopy. (14−18)

Supporting Information

Click to copy section linkSection link copied!

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jcim.3c01278.

  • Histograms of the rotatable bonds and conformers in the COD set, comparisons of RMSD and radius of gyration for the COD and Platinum sets, and histograms of torsional angle deviations for GFN2-optimized and ωB97X-D3/def2-SVP-optimized geometries (PDF)

Terms & Conditions

Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.

Author Information

Click to copy section linkSection link copied!
  • Corresponding Author
    • Geoffrey R. Hutchison - Department of Chemistry, University of Pittsburgh, 219 Parkman Avenue, Pittsburgh, Pennsylvania 15260, United States;  Department of Chemical & Petroleum Engineering, University of Pittsburgh, 3700 O’Hara Street, Pittsburgh, Pennsylvania 15261, United States;  Orcidhttps://orcid.org/0000-0002-1757-1980;  Email: [email protected]
  • Authors
    • Dakota L. Folmsbee - Department of Chemistry, University of Pittsburgh, 219 Parkman Avenue, Pittsburgh, Pennsylvania 15260, United States;  Department of Anesthesiology & Perioperative Medicine, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15261, United States
    • David R. Koes - Department of Computational & Systems Biology, School of Medicine, University of Pittsburgh, Pittsburgh, Pennsylvania 15260, United States;  Orcidhttps://orcid.org/0000-0002-6892-6614
  • Notes
    The authors declare no competing financial interest.

    Raw data, Python notebooks, and torsion pattern figures can be found at https://github.com/hutchisonlab/quantum-torsions. All optimized geometries from CREST/GFN2 are available on Figshare https://doi.org/10.6084/m9.figshare.21395061.v1.

Acknowledgments

Click to copy section linkSection link copied!

We acknowledge the National Science Foundation (CHE-1800435 and CHE-2102474) for support and the University of Pittsburgh Center for Research Computing for the resources provided. Specifically, this work used the H2P cluster, which is supported by NSF award OAC-2117681, as well as resources provided by the Open Science Grid, (72,73) which is supported by NSF award OAC-2030508 and the U.S. Department of Energy’s Office of Science. We also thank Sereina Riniker and Greg Landrum for helpful discussions.

References

Click to copy section linkSection link copied!

This article references 73 other publications.

  1. 1
    Hawkins, P. C. D. Conformation Generation: The State of the Art. J. Chem. Inf. Model. 2017, 57, 1747– 1756,  DOI: 10.1021/acs.jcim.7b00221
    Google Scholar
    1
    Conformation Generation: The State of the Art
    Hawkins, Paul C. D.
    Journal of Chemical Information and Modeling (2017), 57 (8), 1747-1756CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The generation of conformations for small mols. is a problem of continuing interest in cheminformatics and computational drug discovery. This review will present an overview of methods used to sample conformational space, focusing on those methods designed for org. mols. commonly of interest in drug discovery. Different approaches to both the sampling of conformational space and the scoring of conformational stability will be compared and contrasted, with an emphasis on those methods suitable for conformer sampling of large nos. of drug-like mols. Particular attention will be devoted to the appropriate utilization of information from exptl. solid-state structures in validating and evaluating the performance of these tools. The review will conclude with some areas worthy of further investigation.
  2. 2
    Friedrich, N.-O.; de Bruyn Kops, C.; Flachsenberg, F.; Sommer, K.; Rarey, M.; Kirchmair, J. Benchmarking Commercial Conformer Ensemble Generators. J. Chem. Inf. Model. 2017, 57, 2719– 2728,  DOI: 10.1021/acs.jcim.7b00505
    Google Scholar
    2
    Benchmarking Commercial Conformer Ensemble Generators
    Friedrich, Nils-Ole; de Bruyn Kops, Christina; Flachsenberg, Florian; Sommer, Kai; Rarey, Matthias; Kirchmair, Johannes
    Journal of Chemical Information and Modeling (2017), 57 (11), 2719-2728CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    We assess and compare the performance of eight com. conformer ensemble generators (ConfGen, ConfGenX, cxcalc, iCon, MOE LowModeMD, MOE Stochastic, MOE Conformation Import, and OMEGA) and one leading free algorithm, the distance geometry algorithm implemented in RDKit. The comparative study is based on a new version of the Platinum Diverse Dataset, a high-quality benchmarking dataset of 2859 protein-bound ligand conformations extd. from the PDB. Differences in the performance of com. algorithms are much smaller than those obsd. for free algorithms in our previous study. For com. algorithms, the median min. root-mean-square deviations measured between protein-bound ligand conformations and ensembles of a max. of 250 conformers are between 0.46 and 0.61 Å. Com. conformer ensemble generators are characterized by their high robustness, with at least 99% of all input mols. successfully processed and few or even no substantial geometrical errors detectable in their output conformations. The RDKit distance geometry algorithm (with minimization enabled) appears to be a good free alternative since its performance is comparable to that of the midranked com. algorithms. Based on a statistical anal., we elaborate on which algorithms to use and how to parametrize them for best performance in different application scenarios.
  3. 3
    Friedrich, N.-O.; Meyder, A.; de Bruyn Kops, C.; Sommer, K.; Flachsenberg, F.; Rarey, M.; Kirchmair, J. High-Quality Dataset of Protein-Bound Ligand Conformations and Its Application to Benchmarking Conformer Ensemble Generators. J. Chem. Inf. Model. 2017, 57, 529– 539,  DOI: 10.1021/acs.jcim.6b00613
    Google Scholar
    3
    High-Quality Dataset of Protein-Bound Ligand Conformations and Its Application to Benchmarking Conformer Ensemble Generators
    Friedrich, Nils-Ole; Meyder, Agnes; de Bruyn Kops, Christina; Sommer, Kai; Flachsenberg, Florian; Rarey, Matthias; Kirchmair, Johannes
    Journal of Chemical Information and Modeling (2017), 57 (3), 529-539CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The authors developed a cheminformatics pipeline for the fully automated selection and extn. of high-quality protein-bound ligand conformations from x-ray structural data. The pipeline evaluates the validity and accuracy of the 3D structures of small mols. according to multiple criteria, including their fit to the electron d. and their physicochem. and structural properties. Using this approach, the authors compiled two high-quality datasets from the Protein Data Bank (PDB): a comprehensive dataset and a diversified subset of 4626 and 2912 structures, resp. The datasets were applied to benchmarking seven freely available conformer ensemble generators: Balloon (two different algorithms), the RDKit std. conformer ensemble generator, the Exptl.-Torsion basic Knowledge Distance Geometry (ETKDG) algorithm, Confab, Frog2 and Multiconf-DOCK. Substantial differences in the performance of the individual algorithms were obsd., with RDKit and ETKDG generally achieving a favorable balance of accuracy, ensemble size and runtime. The platinum datasets are available for download from http://www.zbh.uni-hamburg.de/platinum_dataset.
  4. 4
    Ebejer, J.-P.; Morris, G. M.; Deane, C. M. Freely Available Conformer Generation Methods: How Good Are They?. J. Chem. Inf. Model. 2012, 52, 1146– 1158,  DOI: 10.1021/ci2004658
    Google Scholar
    4
    Freely Available Conformer Generation Methods: How Good Are They?
    Ebejer, Jean-Paul; Morris, Garrett M.; Deane, Charlotte M.
    Journal of Chemical Information and Modeling (2012), 52 (5), 1146-1158CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    A review. Conformer generation has important implications in cheminformatics, particularly in computational drug discovery where the quality of conformer generation software may affect the outcome of a virtual screening exercise. We examine the performance of four freely available small mol. conformer generation tools (Balloon, Confab, Frog2, and RDKit) alongside a com. tool (MOE). The aim of this study is 3-fold: (i) to identify which tools most accurately reproduce exptl. detd. structures; (ii) to examine the diversity of the generated conformational set; and (iii) to benchmark the computational time expended. These aspects were tested using a set of 708 drug-like mols. assembled from the OMEGA validation set and the Astex Diverse Set. These mols. have varying physicochem. properties and at least one known X-ray crystal structure. We found that RDKit and Confab are statistically better than other methods at generating low rmsd conformers to the known structure. RDKit is particularly suited for less flexible mols. while Confab, with its systematic approach, is able to generate conformers which are geometrically closer to the exptl. detd. structure for mols. with a large no. of rotatable bonds (≥10). In our tests RDKit also resulted as the second fastest method after Frog2. In order to enhance the performance of RDKit, we developed a postprocessing algorithm to build a diverse and representative set of conformers which also contains a close conformer to the known structure. Our anal. indicates that, with postprocessing, RDKit is a valid free alternative to com., proprietary software.
  5. 5
    Penner, P.; Guba, W.; Schmidt, R.; Meyder, A.; Stahl, M.; Rarey, M. The Torsion Library: Semiautomated Improvement of Torsion Rules with SMARTScompare. J. Chem. Inf. Model. 2022, 62, 1644– 1653,  DOI: 10.1021/acs.jcim.2c00043
    Google Scholar
    5
    The Torsion Library: Semiautomated Improvement of Torsion Rules with SMARTScompare
    Penner, Patrick; Guba, Wolfgang; Schmidt, Robert; Meyder, Agnes; Stahl, Martin; Rarey, Matthias
    Journal of Chemical Information and Modeling (2022), 62 (7), 1644-1653CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The Torsion Library is a collection of torsion motifs assocd. with angle distributions, derived from crystallog. databases. It is used in strain assessment, conformer generation, and geometry optimization. A hierarchical structure of expert curated SMARTS defines the chem. environments of rotatable bonds and assocs. these with preferred angles. SMARTS can be very complex and full of implications, which make them difficult to maintain manually. Recent developments in automatically comparing SMARTS patterns can be applied to the Torsion Library to ensure its correctness. We specifically discuss the implementation and the limits of such a procedure in the context of torsion motifs and show several examples of how the Torsion Library benefits from this. All automated changes are validated manually and then shown to have an effect on the angle distributions by correcting matching behavior. The cor. Torsion Library itself is available including both PDB as well as CSD histograms in the Supporting Information and can be used to evaluate rotatable bonds at https://torsions.zbh.uni-hamburg.de.
  6. 6
    Wang, S.; Witek, J.; Landrum, G. A.; Riniker, S. Improving Conformer Generation for Small Rings and Macrocycles Based on Distance Geometry and Experimental Torsional-Angle Preferences. J. Chem. Inf. Model. 2020, 60, 2044– 2058,  DOI: 10.1021/acs.jcim.0c00025
    Google Scholar
    6
    Improving Conformer Generation for Small Rings and Macrocycles Based on Distance Geometry and Experimental Torsional-Angle Preferences
    Wang, Shuzhe; Witek, Jagna; Landrum, Gregory A.; Riniker, Sereina
    Journal of Chemical Information and Modeling (2020), 60 (4), 2044-2058CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The conformer generator ETKDG is a stochastic search method that utilizes distance geometry together with knowledge derived from exptl. crystal structures. It has been shown to generate good conformers for acyclic, flexible mols. This work builds on ETKDG to improve conformer generation of mols. contg. small or large aliph. (i.e., non-arom.) rings. For one, we devise addnl. torsional-angle potentials to describe small aliph. rings and adapt the previously developed potentials for acyclic bonds to facilitate the sampling of macrocycles. However, due to the larger no. of degrees of freedom of macrocycles, the conformational space to sample is much broader than for small mols., creating a challenge for conformer generators. We therefore introduce different heuristics to restrict the search space of macrocycles and bias the sampling toward more exptl. relevant structures. Specifically, we show the usage of elliptical geometry and customizable Coulombic interactions as heuristics. The performance of the improved ETKDG is demonstrated on test sets of diverse macrocycles and cyclic peptides. The code developed here will be incorporated into the 2020.03 release of the open-source cheminformatics library RDKit.
  7. 7
    Riniker, S.; Landrum, G. A. Better Informed Distance Geometry: Using What We Know To Improve Conformation Generation. J. Chem. Inf. Model. 2015, 55, 2562– 2574,  DOI: 10.1021/acs.jcim.5b00654
    Google Scholar
    7
    Better Informed Distance Geometry: Using What We Know To Improve Conformation Generation
    Riniker, Sereina; Landrum, Gregory A.
    Journal of Chemical Information and Modeling (2015), 55 (12), 2562-2574CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    Small org. mols. are often flexible, i.e., they can adopt a variety of low-energy conformations in soln. that exist in equil. with each other. Two main search strategies are used to generate representative conformational ensembles for mols.: systematic and stochastic. In the first approach, each rotatable bond is sampled systematically in discrete intervals, limiting its use to mols. with a small no. of rotatable bonds. Stochastic methods, however, sample the conformational space of a mol. randomly and can thus be applied to more flexible mols. Different methods employ different degrees of exptl. data for conformer generation. So-called knowledge-based methods use predefined libraries of torsional angles and ring conformations. In the distance geometry approach, however, a smaller amt. of empirical information was used, i.e., ideal bond lengths, ideal bond angles, and a few ideal torsional angles. Distance geometry is a computationally fast method to generate conformers, but it has the downside that purely distance-based constraints tend to lead to distorted arom. rings and sp2 centers. To correct this, the resulting conformations are often minimized with a force field, adding computational complexity and run time. Here the authors present an alternative strategy that combines the distance geometry approach with exptl. torsion-angle preferences obtained from small-mol. crystallog. data. The torsional angles are described by a previously developed set of hierarchically structured SMARTS patterns. The new approach is implemented in the open-source cheminformatics library RDKit, and its performance is assessed by comparing the diversity of the generated ensemble and the ability to reproduce crystal conformations taken from the crystal structures of small mols. and protein-ligand complexes.
  8. 8
    Guba, W.; Meyder, A.; Rarey, M.; Hert, J. Torsion Library Reloaded: A New Version of Expert-Derived SMARTS Rules for Assessing Conformations of Small Molecules. J. Chem. Inf. Model. 2016, 56, 1– 5,  DOI: 10.1021/acs.jcim.5b00522
    Google Scholar
    8
    Torsion Library Reloaded: A New Version of Expert-Derived SMARTS Rules for Assessing Conformations of Small Molecules
    Guba, Wolfgang; Meyder, Agnes; Rarey, Matthias; Hert, Jerome
    Journal of Chemical Information and Modeling (2016), 56 (1), 1-5CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The Torsion Library contains hundreds of rules for small mol. conformations which have been derived from the Cambridge Structural Database (CSD) and are curated by mol. design experts. The torsion rules are encoded as SMARTS patterns and categorize rotatable bonds via a traffic light coloring scheme. We have systematically revised all torsion rules to better identify highly strained conformations and minimize the no. of false alerts for CSD small mol. X-ray structures. For this new release, we added or substantially modified 78 torsion patterns and reviewed all angles and tolerance intervals. The overall no. of red alerts for a filtered CSD data set with 130 000 structures was reduced by a factor of 4 compared to the predecessor. This is of clear advantage in 3D virtual screening where hits should only be removed by a conformational filter if they are in energetically inaccessible conformations.
  9. 9
    Gražulis, S.; Chateigner, D.; Downs, R. T.; Yokochi, A. F. T.; Quirós, M.; Lutterotti, L.; Manakova, E.; Butkus, J.; Moeck, P.; Le Bail, A. Crystallography Open Database – an open-access collection of crystal structures. J. Appl. Crystallogr. 2009, 42, 726– 729,  DOI: 10.1107/S0021889809016690
    Google Scholar
    9
    Crystallography Open Database - an open-access collection of crystal structures
    Grazulis, Saulius; Chateigner, Daniel; Downs, Robert T.; Yokochi, A. F. T.; Quiros, Miguel; Lutterotti, Luca; Manakova, Elena; Butkus, Justas; Moeck, Peter; Le Bail, Armel
    Journal of Applied Crystallography (2009), 42 (4), 726-729CODEN: JACGAR; ISSN:0021-8898. (International Union of Crystallography)
    The Crystallog. Open Database (COD), which is a project that aims to gather all available inorg., metal-org. and small org. mol. structural data in one database, is described. The database adopts an open-access model. The COD currently contains ∼80,000 entries in crystallog. information file format, with nearly full coverage of the International Union of Crystallog. publications, and is growing in size and quality.
  10. 10
    Gražulis, S.; Daškevič, A.; Merkys, A.; Chateigner, D.; Lutterotti, L.; Quirós, M.; Serebryanaya, N. R.; Moeck, P.; Downs, R. T.; Le Bail, A. Crystallography Open Database (COD): an open-access collection of crystal structures and platform for world-wide collaboration. Nucleic Acids Res. 2012, 40, D420– D427,  DOI: 10.1093/nar/gkr900
    Google Scholar
    10
    Crystallography Open Database (COD): an open-access collection of crystal structures and platform for world-wide collaboration
    Grazulis, Saulius; Daskevic, Adriana; Merkys, Andrius; Chateigner, Daniel; Lutterotti, Luca; Quiros, Miguel; Serebryanaya, Nadezhda R.; Moeck, Peter; Downs, Robert T.; Le Bail, Armel
    Nucleic Acids Research (2012), 40 (D1), D420-D427CODEN: NARHAD; ISSN:0305-1048. (Oxford University Press)
    Using an open-access distribution model, the Crystallog. Open Database (COD, http://www.crystallog.net) collects all known small mol. / small to medium sized unit cell' crystal structures and makes them available freely on the Internet. As of today, the COD has aggregated ∼150 000 structures, offering basic search capabilities and the possibility to download the whole database, or parts thereof using a variety of std. open communication protocols. A newly developed website provides capabilities for all registered users to deposit published and so far unpublished structures as personal communications or pre-publication depositions. Such a setup enables extension of the COD database by many users simultaneously. This increases the possibilities for growth of the COD database, and is the first step towards establishing a world wide Internet-based collaborative platform dedicated to the collection and curation of structural knowledge.
  11. 11
    Groom, C. R.; Bruno, I. J.; Lightfoot, M. P.; Ward, S. C. The Cambridge Structural Database. Acta Crystallogr. B 2016, 72, 171– 179,  DOI: 10.1107/S2052520616003954
    Google Scholar
    11
    The Cambridge Structural Database
    Groom, Colin R.; Bruno, Ian J.; Lightfoot, Matthew P.; Ward, Suzanna C.
    Acta Crystallographica, Section B: Structural Science, Crystal Engineering and Materials (2016), 72 (2), 171-179CODEN: ACSBDA; ISSN:2052-5206. (International Union of Crystallography)
    The Cambridge Structural Database (CSD) contains a complete record of all published org. and metal-org. small-mol. crystal structures. The database has been in operation for over 50 years and continues to be the primary means of sharing structural chem. data and knowledge across disciplines. As well as structures that are made public to support scientific articles, it includes many structures published directly as CSD Communications. All structures are processed both computationally and by expert structural chem. editors prior to entering the database. A key component of this processing is the reliable assocn. of the chem. identity of the structure studied with the exptl. data. This important step helps ensure that data is widely discoverable and readily reusable. Content is further enriched through selective inclusion of addnl. exptl. data. Entries are available to anyone through free CSD community web services. Linking services developed and maintained by the CCDC, combined with the use of std. identifiers, facilitate discovery from other resources. Data can also be accessed through CCDC and third party software applications and through an application programming interface.
  12. 12
    Sadowski, P.; Baldi, P. Small-Molecule 3D Structure Prediction Using Open Crystallography Data. J. Chem. Inf. Model. 2013, 53, 3127– 3130,  DOI: 10.1021/ci4005282
    Google Scholar
    12
    Small-Molecule 3D Structure Prediction Using Open Crystallography Data
    Sadowski, Peter; Baldi, Pierre
    Journal of Chemical Information and Modeling (2013), 53 (12), 3127-3130CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    Predicting the 3D structures of small mols. is a common problem in chemoinformatics. Even the best methods are inaccurate for complex mols., and there is a large gap in accuracy between proprietary and free algorithms. Previous work presented COSMOS, a novel data-driven algorithm that uses knowledge of known structures from the Cambridge Structural Database and demonstrates performance that was competitive with proprietary algorithms. However, dependence on the Cambridge Structural Database prevented its widespread use. Here, we present an updated version of the COSMOS structure predictor, complete with a free structure library derived from open data sources. We demonstrate that COSMOS performs better than other freely available methods, with a mean RMSD of 1.16 and 1.68 Å for org. and metal-org. structures, resp., and a mean prediction time of 60 ms per mol. This is a 17% and 20% redn., resp., in RMSD compared to the free predictor provided by Open Babel, and it is 10 times faster. The ChemDB Web portal provides a COSMOS prediction Web server, as well as downloadable copies of the COSMOS executable and library of mol. substructures.
  13. 13
    Wicker, J. G. P.; Cooper, R. I. Beyond Rotatable Bond Counts: Capturing 3D Conformational Flexibility in a Single Descriptor. J. Chem. Inf. Model. 2016, 56, 2347– 2352,  DOI: 10.1021/acs.jcim.6b00565
    Google Scholar
    13
    Beyond Rotatable Bond Counts: Capturing 3D Conformational Flexibility in a Single Descriptor
    Wicker, Jerome G. P.; Cooper, Richard I.
    Journal of Chemical Information and Modeling (2016), 56 (12), 2347-2352CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    A new mol. descriptor, nConf20, based on chem. connectivity, is presented which captures the accessible conformational space of a mol. Currently the best available two-dimensional descriptors for quantifying the flexibility of a particular mol. are the rotatable bond count (RBC) and the Kier flexibility index. The authors present a descriptor which captures this information by sampling the conformational space of a mol. using the RDKit conformer generator. Flexibility has previously been identified as a key feature in detg. whether a mol. is likely to crystallize or not. For this application, nConf20 significantly outperforms previously reported single-variable classifiers and also assists rule-based anal. of black-box machine learning classification algorithms.
  14. 14
    Das, S.; Dinpazhoh, L.; Tanemura, K. A.; Merz, K. M. Rapid and Automated Ab Initio Metabolite Collisional Cross Section Prediction from SMILES Input. J. Chem. Inf. Model. 2023, 63, 4995– 5000,  DOI: 10.1021/acs.jcim.3c00890
    Google Scholar
    14
    Rapid and Automated Ab Initio Metabolite Collisional Cross Section Prediction from SMILES Input
    Das, Susanta; Dinpazhoh, Laleh; Tanemura, Kiyoto Aramis; Merz Jr., Kenneth M.
    Journal of Chemical Information and Modeling (2023), 63 (16), 4995-5000CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    We implemented an ab initio CCS prediction workflow which incrementally refines generated structures using mol. mechanics, a deep learning potential, conformational clustering, and quantum mechanics (QM). Automating intermediate steps for a high performance computing (HPC) environment allows users to input the SMILES structure of small org. mols. and obtain a Boltzmann averaged collisional cross section (CCS) value as output. The CCS of a mol. species is a metric measured by ion mobility spectrometry (IMS) which can improve annotation of untargeted metabolomics expts. We report only a minor drop in accuracy when we expedite the CCS calcn. by replacing the QM geometry refinement step with a single-point energy calcn. Even though the workflow involves stochastic steps (i.e., conformation generation and clustering), the final CCS value was highly reproducible for multiple iterations on L-carnosine. Finally, we illustrate that the gas phase ensemble modeled for the workflow are intermediate files which can be used for the prediction of other properties such as aq. phase NMR chem. shift prediction.
  15. 15
    Das, S.; Tanemura, K. A.; Dinpazhoh, L.; Keng, M.; Schumm, C.; Leahy, L.; Asef, C. K.; Rainey, M.; Edison, A. S.; Fernández, F. M.; Merz, K. M. In Silico Collision Cross Section Calculations to Aid Metabolite Annotation. J. Am. Soc. Mass Spectrom. 2022, 33, 750– 759,  DOI: 10.1021/jasms.1c00315
    Google Scholar
    15
    In Silico Collision Cross Section Calculations to Aid Metabolite Annotation
    Das, Susanta; Tanemura, Kiyoto Aramis; Dinpazhoh, Laleh; Keng, Mithony; Schumm, Christina; Leahy, Lydia; Asef, Carter K.; Rainey, Markace; Edison, Arthur S.; Fernandez, Facundo M.; Merz Jr., Kenneth M.
    Journal of the American Society for Mass Spectrometry (2022), 33 (5), 750-759CODEN: JAMSEF; ISSN:1879-1123. (American Chemical Society)
    The interpretation of ion mobility coupled to mass spectrometry (IM-MS) data to predict unknown structures is challenging and depends on accurate theor. ests. of the mol. ion collision cross section (CCS) against a buffer gas in a low or atm. pressure drift chamber. The sensitivity and reliability of computational prediction of CCS values depend on accurately modeling the mol. state over accessible conformations. In this work, we developed an efficient CCS computational workflow using a machine learning model in conjunction with std. DFT methods and CCS calcns. Furthermore, we have performed Traveling Wave IM-MS (TWIMS) expts. to validate the extant exptl. values and assess uncertainties in exptl. measured CCS values. The developed workflow yielded accurate structural predictions and provides unique insights into the likely preferred conformation analyzed using IM-MS expts. The complete workflow makes the computation of CCS values tractable for a large no. of conformationally flexible metabolites with complex mol. structures.
  16. 16
    Insausti, A.; Alonso, E. R.; Tercero, B.; Santos, J. I.; Calabrese, C.; Vogt, N.; Corzana, F.; Demaison, J.; Cernicharo, J.; Cocinero, E. J. Laboratory Observation of, Astrochemical Search for, and Structure of Elusive Erythrulose in the Interstellar Medium. J. Phys. Chem. Lett. 2021, 12, 1352– 1359,  DOI: 10.1021/acs.jpclett.0c03050
    Google Scholar
    16
    Laboratory Observation of, Astrochemical Search for, and Structure of Elusive Erythrulose in the Interstellar Medium
    Insausti, Aran; Alonso, Elena R.; Tercero, Belen; Santos, Jose I.; Calabrese, Camilla; Vogt, Natalja; Corzana, Francisco; Demaison, Jean; Cernicharo, Jose; Cocinero, Emilio J.
    Journal of Physical Chemistry Letters (2021), 12 (4), 1352-1359CODEN: JPCLCD; ISSN:1948-7185. (American Chemical Society)
    Rotational spectroscopy provides the most powerful means of identifying mols. of biol. interest in the interstellar medium (ISM), but despite their importance, the detection of carbohydrates has remained rather elusive. Here, we present a comprehensive Fourier transform rotational spectroscopic study of elusive erythrulose, a sugar building block likely to be present in the ISM, employing a novel method of transferring the hygroscopic oily carbohydrate into the gas phase. The high sensitivity of the expt. allowed the rotational spectra of all monosubstituted isotopologue species of 13C-12C3H8O4 to be recorded, which, together with quantum chem. calcns., enabled us to det. their equil. geometries (reSE) with great precision. Searches employing the new exptl. data for erythrulose have been undertaken in different ISM regions, so far including the cold areas Barnard 1, the pre-stellar core TMC-1, Sagittarius B2. Although no lines of erythrulose were found, this data will serve to enable future searches and possible detections in other ISM regions.
  17. 17
    Alonso, E. R.; Peña, I.; Cabezas, C.; Alonso, J. L. Structural Expression of Exo-Anomeric Effect. J. Phys. Chem. Lett. 2016, 7, 845– 850,  DOI: 10.1021/acs.jpclett.6b00028
    Google Scholar
    17
    Structural Expression of Exo-Anomeric Effect
    Alonso, Elena R.; Pena, Isabel; Cabezas, Carlos; Alonso, Jose L.
    Journal of Physical Chemistry Letters (2016), 7 (5), 845-850CODEN: JPCLCD; ISSN:1948-7185. (American Chemical Society)
    Structural signatures for exo-anomeric effect have been extd. from the archetypal Me β-D-xyloside using broadband Fourier transform microwave spectroscopy combined with laser ablation. Spectrum anal. allows the detn. of a set of rotational consts., which has been unequivocally attributed to conformer cc-β-4C1 g-, corresponding to the global min. of the potential energy surface, where the aglycon residue (CH3) orientation contributes to maximization of the exo-anomeric effect. Further anal. allowed the detn. of the rs structure, based on the detection of 11 isotopologues-derived from the presence of six 13C and five 18O atoms-obsd. in their natural abundances. The obsd. glycosidic C1-O1 bond length decrease (1.38 Å) can be interpreted in terms of the exo-anomeric effect. As such, the exo-anomeric effect presents itself as one of the main driving forces controlling the shape of many biol. important oligosaccharides.
  18. 18
    Peña, I.; Cocinero, E. J.; Cabezas, C.; Lesarri, A.; Mata, S.; Écija, P.; Daly, A. M.; Cimas, A.; Bermúdez, C.; Basterretxea, F. J.; Blanco, S.; Fernández, J. A.; López, J. C.; Castaño, F.; Alonso, J. L. Six Pyranoside Forms of Free 2-Deoxy-D-ribose. Angew. Chem., Int. Ed. 2013, 52, 11840– 11845,  DOI: 10.1002/anie.201305589
    Google Scholar
    18
    Six Pyranoside Forms of Free 2-Deoxy-D-ribose
    Pena, Isabel; Cocinero, Emilio J.; Cabezas, Carlos; Lesarri, Alberto; Mata, Santiago; Ecija, Patricia; Daly, Adam M.; Cimas, Alvaro; Bermudez, Celina; Basterretxea, Francisco J.; Blanco, Susana; Fernandez, Jose A.; Lopez, Juan C.; Castano, Fernando; Alonso, Jose L.
    Angewandte Chemie, International Edition (2013), 52 (45), 11840-11845CODEN: ACIEF5; ISSN:1433-7851. (Wiley-VCH Verlag GmbH & Co. KGaA)
    The gas phase rotational spectrum of 2-deoxy-D-ribose (I) was obsd. using a UV ultrafast laser ablation technique as a source for Balle-Flygare and chirped pulsed FTMW spectrometers and was assigned with the aide of MP2(FULL) calcns. In the gas phase, I exists as a mixt. of two α- (10%) and four β-pyranose (90%) conformers. the conformational behavior is controlled by both anomeric effects and hydrogen bonding. the OH groups are preferentially oriented so as to favor cooperative hydrogen bonding. In this context, the previous exptl. ionization potential of 9.1 eV of gas phase I (obtained using tunable vacuum UV synchrotron radiation ) assigned to the α-pyranose should actually correspond to the β-pyranose. the solvent effect on furanose-pyranose equil. and the occurrence of these conformers in RNA and DNA were discussed.
  19. 19
    Baldi, P. Data-Driven High-Throughput Prediction of the 3-D Structure of Small Molecules: Review and Progress. A Response to the Letter by the Cambridge Crystallographic Data Centre. J. Chem. Inf. Model. 2011, 51, 3029,  DOI: 10.1021/ci200460z
    Google Scholar
    19
    Data-Driven High-Throughput Prediction of the 3-D Structure of Small Molecules: Review and Progress. A Response to the Letter by the Cambridge Crystallographic Data Centre
    Baldi, Pierre
    Journal of Chemical Information and Modeling (2011), 51 (12), 3029CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    A polemic in response to Groom, C.R. (ibid., S1, 2787 - 2787, 2011) is presented to sentiments expressed in "Data-Driven High-Throughput Prediction of the 3-D Structure of Small Mols.: Review and Progress. A Response from The Cambridge Crystallog. Data Center", recently published in the Journal of Chem. Information and Modeling, which may give readers a misleading impression regarding significant impediments to scientific research posed by the CCDC.
  20. 20
    Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.; Skiff, W. M. UFF, a full periodic table force field for molecular mechanics and molecular dynamics simulations. J. Am. Chem. Soc. 1992, 114, 10024– 10035,  DOI: 10.1021/ja00051a040
    Google Scholar
    20
    UFF, a full periodic table force field for molecular mechanics and molecular dynamics simulations
    Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A., III; Skiff, W. M.
    Journal of the American Chemical Society (1992), 114 (25), 10024-35CODEN: JACSAT; ISSN:0002-7863.
    A new mol. mechanics force field, the Universal force field (UFF), is described wherein the force field parameters are estd. using general rules based only on the element, its hybridization and its connectivity. The force field functional forms, parameters, and generating formulas for the full periodic table are presented.
  21. 21
    Folmsbee, D.; Hutchison, G. Assessing conformer energies using electronic structure and machine learning methods. Int. J. Quantum Chem. 2021, 121, e26381  DOI: 10.1002/qua.26381
    Google Scholar
    21
    Assessing conformer energies using electronic structure and machine learning methods
    Folmsbee, Dakota; Hutchison, Geoffrey
    International Journal of Quantum Chemistry (2021), 121 (1), e26381CODEN: IJQCB2; ISSN:0020-7608. (John Wiley & Sons, Inc.)
    A review. We have performed a large-scale evaluation of current computational methods, including conventional small-mol. force fields; semiempirical, d. functional, ab initio electronic structure methods; and current machine learning (ML) techniques to evaluate relative single-point energies. Using up to 10 local min. geometries across ∼700 mols., each optimized by B3LYP-D3BJ with single-point DLPNO-CCSD(T) triple-zeta energies, we consider over 6500 single points to compare the correlation between different methods for both relative energies and ordered rankings of min. We find that the current ML methods have potential and recommend methods at each tier of the accuracy-time tradeoff, particularly the recent GFN2 semiempirical method, the B97-3c d. functional approxn., and RI-MP2 for accurate conformer energies. The ANI family of ML methods shows promise, particularly the ANI-1ccx variant trained in part on coupled-cluster energies. Multiple methods suggest continued improvements should be expected in both performance and accuracy.
  22. 22
    Kanal, I. Y.; Keith, J. A.; Hutchison, G. R. A sobering assessment of small-molecule force field methods for low energy conformer predictions. Int. J. Quantum Chem. 2018, 118, e25512  DOI: 10.1002/qua.25512
    Google Scholar
    There is no corresponding record for this reference.
  23. 23
    Bannwarth, C.; Ehlert, S.; Grimme, S. GFN2-xTB - An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions. J. Chem. Theory Comput. 2019, 15, 1652– 1671,  DOI: 10.1021/acs.jctc.8b01176
    Google Scholar
    23
    GFN2-xTB-An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions
    Bannwarth, Christoph; Ehlert, Sebastian; Grimme, Stefan
    Journal of Chemical Theory and Computation (2019), 15 (3), 1652-1671CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
    An extended semiempirical tight-binding model is presented, which is primarily designed for the fast calcn. of structures and noncovalent interactions energies for mol. systems with roughly 1000 atoms. The essential novelty in this so-called GFN2-xTB method is the inclusion of anisotropic second order d. fluctuation effects via short-range damped interactions of cumulative at. multipole moments. Without noticeable increase in the computational demands, this results in a less empirical and overall more phys. sound method, which does not require any classical halogen or hydrogen bonding corrections and which relies solely on global and element-specific parameters (available up to radon, Z = 86). Moreover, the at. partial charge dependent D4 London dispersion model is incorporated self-consistently, which can be naturally obtained in a tight-binding picture from second order d. fluctuations. Fully anal. and numerically precise gradients (nuclear forces) are implemented. The accuracy of the method is benchmarked for a wide variety of systems and compared with other semiempirical methods. Along with excellent performance for the "target" properties, we also find lower errors for "off-target" properties such as barrier heights and mol. dipole moments. High computational efficiency along with the improved physics compared to it precursor GFN-xTB makes this method well-suited to explore the conformational space of mol. systems. Significant improvements are furthermore obsd. for various benchmark sets, which are prototypical for biomol. systems in aq. soln.
  24. 24
    Smith, J. S.; Isayev, O.; Roitberg, A. E. ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost. Chem. Sci. 2017, 8, 3192– 3203,  DOI: 10.1039/C6SC05720A
    Google Scholar
    24
    ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost
    Smith, J. S.; Isayev, O.; Roitberg, A. E.
    Chemical Science (2017), 8 (4), 3192-3203CODEN: CSHCCN; ISSN:2041-6520. (Royal Society of Chemistry)
    A review. Deep learning is revolutionizing many areas of science and technol., esp. image, text, and speech recognition. In this paper, we demonstrate how a deep neural network (NN) trained on quantum mech. (QM) DFT calcns. can learn an accurate and transferable potential for org. mols. We introduce ANAKIN-ME (Accurate NeurAl networK engINe for Mol. Energies) or ANI for short. ANI is a new method designed with the intent of developing transferable neural network potentials that utilize a highly-modified version of the Behler and Parrinello symmetry functions to build single-atom at. environment vectors (AEV) as a mol. representation. AEVs provide the ability to train neural networks to data that spans both configurational and conformational space, a feat not previously accomplished on this scale. We utilized ANI to build a potential called ANI-1, which was trained on a subset of the GDB databases with up to 8 heavy atoms in order to predict total energies for org. mols. contg. four atom types: H, C, N, and O. To obtain an accelerated but phys. relevant sampling of mol. potential surfaces, we also proposed a Normal Mode Sampling (NMS) method for generating mol. conformations. Through a series of case studies, we show that ANI-1 is chem. accurate compared to ref. DFT calcns. on much larger mol. systems (up to 54 atoms) than those included in the training data set.
  25. 25
    Smith, J. S.; Nebgen, B.; Lubbers, N.; Isayev, O.; Roitberg, A. E. Less is more: Sampling chemical space with active learning. J. Chem. Phys. 2018, 148, 241733,  DOI: 10.1063/1.5023802
    Google Scholar
    25
    Less is more: Sampling chemical space with active learning
    Smith, Justin S.; Nebgen, Ben; Lubbers, Nicholas; Isayev, Olexandr; Roitberg, Adrian E.
    Journal of Chemical Physics (2018), 148 (24), 241733/1-241733/10CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
    The development of accurate and transferable machine learning (ML) potentials for predicting mol. energetics is a challenging task. The process of data generation to train such ML potentials is a task neither well understood nor researched in detail. In this work, we present a fully automated approach for the generation of datasets with the intent of training universal ML potentials. It is based on the concept of active learning (AL) via Query by Committee (QBC), which uses the disagreement between an ensemble of ML potentials to infer the reliability of the ensemble's prediction. QBC allows the presented AL algorithm to automatically sample regions of chem. space where the ML potential fails to accurately predict the potential energy. AL improves the overall fitness of ANAKIN-ME (ANI) deep learning potentials in rigorous test cases by mitigating human biases in deciding what new training data to use. AL also reduces the training set size to a fraction of the data required when using naive random sampling techniques. To provide validation of our AL approach, we develop the COmprehensive Machine-learning Potential (COMP6) benchmark (publicly available on GitHub) which contains a diverse set of org. mols. Active learning-based ANI potentials outperform the original random sampled ANI-1 potential with only 10% of the data, while the final active learning-based model vastly outperforms ANI-1 on the COMP6 benchmark after training to only 25% of the data. Finally, we show that our proposed AL technique develops a universal ANI potential (ANI-1x) that provides accurate energy and force predictions on the entire COMP6 benchmark. This universal ML potential achieves a level of accuracy on par with the best ML potentials for single mols. or materials, while remaining applicable to the general class of org. mols. composed of the elements CHNO. (c) 2018 American Institute of Physics.
  26. 26
    Smith, J. S.; Nebgen, B. T.; Zubatyuk, R.; Lubbers, N.; Devereux, C.; Barros, K.; Tretiak, S.; Isayev, O.; Roitberg, A. E. Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning. Nat. Commun. 2019, 10, 2903,  DOI: 10.1038/s41467-019-10827-4
    Google Scholar
    26
    Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning
    Smith Justin S; Devereux Christian; Roitberg Adrian E; Smith Justin S; Nebgen Benjamin T; Zubatyuk Roman; Lubbers Nicholas; Barros Kipton; Tretiak Sergei; Smith Justin S; Lubbers Nicholas; Nebgen Benjamin T; Tretiak Sergei; Zubatyuk Roman; Isayev Olexandr
    Nature communications (2019), 10 (1), 2903 ISSN:.
    Computational modeling of chemical and biological systems at atomic resolution is a crucial tool in the chemist's toolset. The use of computer simulations requires a balance between cost and accuracy: quantum-mechanical methods provide high accuracy but are computationally expensive and scale poorly to large systems, while classical force fields are cheap and scalable, but lack transferability to new systems. Machine learning can be used to achieve the best of both approaches. Here we train a general-purpose neural network potential (ANI-1ccx) that approaches CCSD(T)/CBS accuracy on benchmarks for reaction thermochemistry, isomerization, and drug-like molecular torsions. This is achieved by training a network to DFT data then using transfer learning techniques to retrain on a dataset of gold standard QM calculations (CCSD(T)/CBS) that optimally spans chemical space. The resulting potential is broadly applicable to materials science, biology, and chemistry, and billions of times faster than CCSD(T)/CBS calculations.
  27. 27
    Devereux, C.; Smith, J. S.; Huddleston, K. K.; Barros, K.; Zubatyuk, R.; Isayev, O.; Roitberg, A. E. Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens. J. Chem. Theory Comput. 2020, 16, 4192– 4202,  DOI: 10.1021/acs.jctc.0c00121
    Google Scholar
    27
    Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens
    Devereux, Christian; Smith, Justin S.; Huddleston, Kate K.; Barros, Kipton; Zubatyuk, Roman; Isayev, Olexandr; Roitberg, Adrian E.
    Journal of Chemical Theory and Computation (2020), 16 (7), 4192-4202CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
    Machine learning (ML) methods have become powerful, predictive tools in a wide range of applications, such as facial recognition and autonomous vehicles. In the sciences, computational chemists and physicists have been using ML for the prediction of phys. phenomena, such as atomistic potential energy surfaces and reaction pathways. Transferable ML potentials, such as ANI-1x, have been developed with the goal of accurately simulating org. mols. contg. the chem. elements H, C, N, and O. Here, we provide an extension of the ANI-1x model. The new model, dubbed ANI-2x, is trained to three addnl. chem. elements: S, F, and Cl. Addnl., ANI-2x underwent torsional refinement training to better predict mol. torsion profiles. These new features open a wide range of new applications within org. chem. and drug development. These seven elements (H, C, N, O, F, Cl, and S) make up ∼ 90% of drug-like mols. To show that these addns. do not sacrifice accuracy, we have tested this model across a range of org. mols. and applications, including the COMP6 benchmark, dihedral rotations, conformer scoring, and nonbonded interactions. ANI-2x is shown to accurately predict mol. energies compared to d. functional theory with a ~ 106 factor speedup and a negligible slowdown compared to ANI-1x and shows subchem. accuracy across most of the COMP6 benchmark. The resulting model is a valuable tool for drug development which can potentially replace both quantum calcns. and classical force fields for a myriad of applications.
  28. 28
    Qiao, Z.; Welborn, M.; Anandkumar, A.; Manby, F. R.; Miller, T. F. OrbNet: Deep learning for quantum chemistry using symmetry-adapted atomic-orbital features. J. Chem. Phys. 2020, 153, 124111,  DOI: 10.1063/5.0021955
    Google Scholar
    28
    OrbNet: Deep learning for quantum chemistry using symmetry-adapted atomic-orbital features
    Qiao, Zhuoran; Welborn, Matthew; Anandkumar, Animashree; Manby, Frederick R.; Miller, Thomas F.
    Journal of Chemical Physics (2020), 153 (12), 124111CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
    We introduce a machine learning method in which energy solns. from the Schroedinger equation are predicted using symmetry adapted AO features and a graph neural-network architecture. OrbNet is shown to outperform existing methods in terms of learning efficiency and transferability for the prediction of d. functional theory results while employing low-cost features that are obtained from semi-empirical electronic structure calcns. For applications to datasets of drug-like mols., including QM7b-T, QM9, GDB-13-T, DrugBank, and the conformer benchmark dataset of Folmsbee and Hutchison [Int. J. Quantum Chem. (published online) (2020)], OrbNet predicts energies within chem. accuracy of d. functional theory at a computational cost that is 1000-fold or more reduced. (c) 2020 American Institute of Physics.
  29. 29
    Christensen, A. S.; Sirumalla, S. K.; Qiao, Z.; O’Connor, M. B.; Smith, D. G. A.; Ding, F.; Bygrave, P. J.; Anandkumar, A.; Welborn, M.; Manby, F. R.; Miller, T. F. OrbNet Denali: A machine learning potential for biological and organic chemistry with semi-empirical cost and DFT accuracy. J. Chem. Phys. 2021, 155, 204103,  DOI: 10.1063/5.0061990
    Google Scholar
    29
    OrbNet Denali: A machine learning potential for biological and organic chemistry with semi-empirical cost and DFT accuracy
    Christensen, Anders S.; Sirumalla, Sai Krishna; Qiao, Zhuoran; O'Connor, Michael B.; Smith, Daniel G. A.; Ding, Feizhi; Bygrave, Peter J.; Anandkumar, Animashree; Welborn, Matthew; Manby, Frederick R.; Miller, Thomas F.
    Journal of Chemical Physics (2021), 155 (20), 204103CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
    We present OrbNet Denali, a machine learning model for an electronic structure that is designed as a drop-in replacement for ground-state d. functional theory (DFT) energy calcns. The model is a message-passing graph neural network that uses symmetry-adapted AO features from a low-cost quantum calcn. to predict the energy of a mol. OrbNet Denali is trained on a vast dataset of 2.3 x 106 DFT calcns. on mols. and geometries. This dataset covers the most common elements in biochem. and org. chem. (H, Li, B, C, N, O, F, Na, Mg, Si, P, S, Cl, K, Ca, Br, and I) and charged mols. OrbNet Denali is demonstrated on several well-established benchmark datasets, and we find that it provides accuracy that is on par with modern DFT methods while offering a speedup of up to three orders of magnitude. For the GMTKN55 benchmark set, OrbNet Denali achieves WTMAD-1 and WTMAD-2 scores of 7.19 and 9.84, on par with modern DFT functionals. For several GMTKN55 subsets, which contain chem. problems that are not present in the training set, OrbNet Denali produces a mean abs. error comparable to those of DFT methods. For the Hutchison conformer benchmark set, OrbNet Denali has a median correlation coeff. of R2 = 0.90 compared to the ref. DLPNO-CCSD(T) calcn. and R2 = 0.97 compared to the method used to generate the training data (ωB97X-D3/def2-TZVP), exceeding the performance of any other method with a similar cost. Similarly, the model reaches chem. accuracy for non-covalent interactions in the S66x10 dataset. For torsional profiles, OrbNet Denali reproduces the torsion profiles of ωB97X-D3/def2-TZVP with an av. mean abs. error of 0.12 kcal/mol for the potential energy surfaces of the diverse fragments in the TorsionNet500 dataset. (c) 2021 American Institute of Physics.
  30. 30
    Nakata, M.; Shimazaki, T. PubChemQC Project: A Large-Scale First-Principles Electronic Structure Database for Data-Driven Chemistry. J. Chem. Inf. Model. 2017, 57, 1300– 1308,  DOI: 10.1021/acs.jcim.7b00083
    Google Scholar
    30
    PubChemQC Project: A Large-Scale First-Principles Electronic Structure Database for Data-Driven Chemistry
    Nakata, Maho; Shimazaki, Tomomi
    Journal of Chemical Information and Modeling (2017), 57 (6), 1300-1308CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    Large-scale mol. databases play an essential role in the investigation of various subjects such as the development of org. materials, in-silico drug designs, and data-driven studies with machine learning, among others. We developed a large-scale quantum chem. database based on the first-principles method without performing any expt. Our database currently contains three million mol. electronic structures based on the d. functional theory method at the B3LYP/6-31G* level, and we successively calcd. 10 low-lying excited states of over two million mols. by the time-dependent DFT method with the 6-31+G* basis set. To select the mols. calcd. in our project, we mainly referred to the PubChem project, and it was used as a source of the mol. structures in short strings using the InChI and the SMILES representations. Accordingly, we named our quantum chem. database project as "PubChemQC" (http://pubchemqc.riken.jp/) and placed it in the public domain. In this paper, we showed the fundamental features of the PubChemQC database and dis- cussed the techniques used to construct the dataset for large-scale quantum chem. calcns. We also presented a machine-learning approach to predict the electronic structure of mols. as an example to demonstrate the suitability of the large-scale quantum chem. database.
  31. 31
    Smith, D. G. A.; Altarawy, D.; Burns, L. A.; Welborn, M.; Naden, L. N.; Ward, L.; Ellis, S.; Pritchard, B. P.; Crawford, T. D. The MolSSI QCArchive project: An open-source platform to compute, organize, and share quantum chemistry data. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2020, 11, e1491  DOI: 10.1002/wcms.1491
    Google Scholar
    There is no corresponding record for this reference.
  32. 32
    Lim, V. T.; Hahn, D. F.; Tresadern, G.; Bayly, C. I.; Mobley, D. L. Benchmark assessment of molecular geometries and energies from small molecule force fields [version 1; peer review: 2 approved]. F1000Research 2020, 9, 1390,  DOI: 10.12688/f1000research.27141.1
    Google Scholar
    32
    Benchmark assessment of molecular geometries and energies from small molecule force fields
    Lim, Victoria T.; Hahn, David F.; Tresadern, Gary; Bayly, Christopher I.; Mobley, David L.
    F1000Research (2020), 9 (), 1390CODEN: FRESJL; ISSN:2046-1402. (F1000 Research Ltd.)
    Background: Force fields are used in a wide variety of contexts for classical mol. simulation, including studies on protein-ligand binding, membrane permeation, and thermophys. property prediction. The quality of these studies relies on the quality of the force fields used to represent the systems. Methods: Focusing on small mols. of fewer than 50 heavy atoms, our aim in this work is to compare nine force fields: GAFF, GAFF2, MMFF94, MMFF94S, OPLS3e, SMIRNOFF99Frosst, and the Open Force Field Parsley, versions 1.0, 1.1, and 1.2. On a dataset comprising 22,675 mol. structures of 3,271 mols., we analyzed force field-optimized geometries and conformer energies compared to ref. quantum mech. (QM) data. Results: We show that while OPLS3e performs best, the latest Open Force Field Parsley release is approaching a comparable level of accuracy in reproducing QM geometries and energetics for this set of mols. Meanwhile, the performance of established force fields such as MMFF94S and GAFF2 is generally somewhat worse. We also fiend that the series of recent Open Force Field versions provide significant increases in accuracy. Conclusions: This study provides an extensive test of the performance of different mol. mechanics force fields on a diverse mol. set, and highlights two (OPLS3e and OpenFF 1.2) that perform better than the others tested on the present comparison. Our mol. set and results are available for other researchers to use in testing.
  33. 33
    Gražulis, S.; Merkys, A.; Vaitkus, A.; Chateigner, D.; Lutterotti, L.; Moeck, P.; Quiros, M.; Downs, R. T.; Kaminsky, W.; Bail, A. L. Materials Informatics: Methods, Tools and Applications; Isayev, O., Tropsha, A., Curtarolo, S., Eds.; Wiley, 2019; Chapter 1, pp 1– 39.
    Google Scholar
    There is no corresponding record for this reference.
  34. 34
    Chai, J.-D.; Head-Gordon, M. Systematic optimization of long-range corrected hybrid density functionals. J. Chem. Phys. 2008, 128, 084106,  DOI: 10.1063/1.2834918
    Google Scholar
    34
    Systematic optimization of long-range corrected hybrid density functionals
    Chai, Jeng-Da; Head-Gordon, Martin
    Journal of Chemical Physics (2008), 128 (8), 084106/1-084106/15CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
    A general scheme for systematically modeling long-range cor. (LC) hybrid d. functionals is proposed. Our resulting two LC hybrid functionals are shown to be accurate in thermochem., kinetics, and noncovalent interactions, when compared with common hybrid d. functionals. The qual. failures of the commonly used hybrid d. functionals in some "difficult problems," such as dissocn. of sym. radical cations and long-range charge-transfer excitations, are significantly reduced by the present LC hybrid d. functionals. (c) 2008 American Institute of Physics.
  35. 35
    Weigend, F.; Ahlrichs, R. Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy. Phys. Chem. Chem. Phys. 2005, 7, 3297,  DOI: 10.1039/b508541a
    Google Scholar
    35
    Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy
    Weigend, Florian; Ahlrichs, Reinhart
    Physical Chemistry Chemical Physics (2005), 7 (18), 3297-3305CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
    Gaussian basis sets of quadruple zeta valence quality for Rb-Rn are presented, as well as bases of split valence and triple zeta valence quality for H-Rn. The latter were obtained by (partly) modifying bases developed previously. A large set of more than 300 mols. representing (nearly) all elements-except lanthanides-in their common oxidn. states was used to assess the quality of the bases all across the periodic table. Quantities investigated were atomization energies, dipole moments and structure parameters for Hartree-Fock, d. functional theory and correlated methods, for which we had chosen Moller-Plesset perturbation theory as an example. Finally recommendations are given which type of basis set is used best for a certain level of theory and a desired quality of results.
  36. 36
    Weigend, F. Accurate Coulomb-fitting basis sets for H to Rn. Phys. Chem. Chem. Phys. 2006, 8, 1057,  DOI: 10.1039/b515623h
    Google Scholar
    36
    Accurate Coulomb-fitting basis sets for H to Rn
    Weigend, Florian
    Physical Chemistry Chemical Physics (2006), 8 (9), 1057-1065CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
    A series of auxiliary basis sets to fit Coulomb potentials for the elements H to Rn (except lanthanides) is presented. For each element only one auxiliary basis set is needed to approx. Coulomb energies in conjunction with orbital basis sets of split valence, triple zeta valence and quadruple zeta valence quality with errors of typically below ca. 0.15 kJ mol-1 per atom; this was demonstrated in conjunction with the recently developed orbital basis sets of types def2-SV(P), def2-TZVP and def2-QZVPP for a large set of small mols. representing (nearly) each element in all of its common oxidn. states. These auxiliary bases are slightly more than three times larger than orbital bases of split valence quality. Compared to non-approximated treatments, computation times for the Coulomb part are reduced by a factor of ca. 8 for def2-SV(P) orbital bases, ca. 25 for def2-TZVP and ca. 100 for def2-QZVPP orbital bases.
  37. 37
    Sterling, T.; Irwin, J. J. ZINC 15 – Ligand Discovery for Everyone. J. Chem. Inf. Model. 2015, 55, 2324– 2337,  DOI: 10.1021/acs.jcim.5b00559
    Google Scholar
    37
    ZINC 15 - Ligand Discovery for Everyone
    Sterling, Teague; Irwin, John J.
    Journal of Chemical Information and Modeling (2015), 55 (11), 2324-2337CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    Many questions about the biol. activity and availability of small mols. remain inaccessible to investigators who could most benefit from their answers. To narrow the gap between chemoinformatics and biol., we have developed a suite of ligand annotation, purchasability, target, and biol. assocn. tools, incorporated into ZINC and meant for investigators who are not computer specialists. The new version contains over 120 million purchasable "drug-like" compds. - effectively all org. mols. that are for sale - a quarter of which are available for immediate delivery. ZINC connects purchasable compds. to high-value ones such as metabolites, drugs, natural products, and annotated compds. from the literature. Compds. may be accessed by the genes for which they are annotated as well as the major and minor target classes to which those genes belong. It offers new anal. tools that are easy for nonspecialists yet with few limitations for experts. ZINC retains its original 3D roots - all mols. are available in biol. relevant, ready-to-dock formats. ZINC is freely available at http://zinc15.docking.org.
  38. 38
    Yoshikawa, N.; Hutchison, G. R. Fast, efficient fragment-based coordinate generation for Open Babel. J. Cheminf. 2019, 11, 49,  DOI: 10.1186/s13321-019-0372-5
    Google Scholar
    There is no corresponding record for this reference.
  39. 39
    O’Boyle, N. M.; Morley, C.; Hutchison, G. R. Pybel: a Python wrapper for the OpenBabel cheminformatics toolkit. Chem. Cent. J. 2008, 2, 5,  DOI: 10.1186/1752-153x-2-5
    Google Scholar
    39
    Pybel: a Python wrapper for the OpenBabel cheminformatics toolkit
    O'Boyle Noel M; Morley Chris; Hutchison Geoffrey R
    Chemistry Central journal (2008), 2 (), 5 ISSN:.
    BACKGROUND: Scripting languages such as Python are ideally suited to common programming tasks in cheminformatics such as data analysis and parsing information from files. However, for reasons of efficiency, cheminformatics toolkits such as the OpenBabel toolkit are often implemented in compiled languages such as C++. We describe Pybel, a Python module that provides access to the OpenBabel toolkit. RESULTS: Pybel wraps the direct toolkit bindings to simplify common tasks such as reading and writing molecular files and calculating fingerprints. Extensive use is made of Python iterators to simplify loops such as that over all the molecules in a file. A Pybel Molecule can be easily interconverted to an OpenBabel OBMol to access those methods or attributes not wrapped by Pybel. CONCLUSION: Pybel allows cheminformaticians to rapidly develop Python scripts that manipulate chemical information. It is open source, available cross-platform, and offers the power of the OpenBabel toolkit to Python programmers.
  40. 40
    O’Boyle, N. M.; Banck, M.; James, C. A.; Morley, C.; Vandermeersch, T.; Hutchison, G. R. Open Babel: An open chemical toolbox. J. Cheminf. 2011, 3, 33,  DOI: 10.1186/1758-2946-3-33
    Google Scholar
    40
    Open Babel: an open chemical toolbox
    O'Boyle, Noel M.; Banck, Michael; James, Craig A.; Morley, Chris; Vandermeersch, Tim; Hutchison, Geoffrey R.
    Journal of Cheminformatics (2011), 3 (), 33CODEN: JCOHB3; ISSN:1758-2946. (Chemistry Central Ltd.)
    Background: A frequent problem in computational modeling is the interconversion of chem. structures between different formats. While std. interchange formats exist (for example, Chem. Markup Language) and de facto stds. have arisen (for example, SMILES format), the need to interconvert formats is a continuing problem due to the multitude of different application areas for chem. data, differences in the data stored by different formats (0D vs. 3D, for example), and competition between software along with a lack of vendor-neutral formats. Results: We discuss, for the first time, Open Babel, an open-source chem. toolbox that speaks the many languages of chem. data. Open Babel version 2.3 interconverts over 110 formats. The need to represent such a wide variety of chem. and mol. data requires a library that implements a wide range of cheminformatics algorithms, from partial charge assignment and aromaticity detection, to bond order perception and canonicalization. We detail the implementation of Open Babel, describe key advances in the 2.3 release, and outline a variety of uses both in terms of software products and scientific research, including applications far beyond simple format interconversion. Conclusions: Open Babel presents a soln. to the proliferation of multiple chem. file formats. In addn., it provides a variety of useful utilities from conformer searching and 2D depiction, to filtering, batch conversion, and substructure and similarity searching. For developers, it can be used as a programming library to handle chem. data in areas such as org. chem., drug design, materials science, and computational chem. It is freely available under an open-source license.
  41. 41
    Wishart, D. S.; Feunang, Y. D.; Guo, A. C.; Lo, E. J.; Marcu, A.; Grant, J. R.; Sajed, T.; Johnson, D.; Li, C.; Sayeeda, Z. DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic Acids Res. 2018, 46, D1074– D1082,  DOI: 10.1093/nar/gkx1037
    Google Scholar
    41
    DrugBank 5.0: a major update to the DrugBank database for 2018
    Wishart, David S.; Feunang, Yannick D.; Guo, An C.; Lo, Elvis J.; Marcu, Ana; Grant, Jason R.; Sajed, Tanvir; Johnson, Daniel; Li, Carin; Sayeeda, Zinat; Assempour, Nazanin; Iynkkaran, Ithayavani; Liu, Yifeng; Maciejewski, Adam; Gale, Nicola; Wilson, Alex; Chin, Lucy; Cummings, Ryan; Le, Diana; Pon, Allison; Knox, Craig; Wilson, Michael
    Nucleic Acids Research (2018), 46 (D1), D1074-D1082CODEN: NARHAD; ISSN:1362-4962. (Oxford University Press)
    DrugBank is a web-enabled database contg. comprehensivemol. information about drugs, their mechanisms, their interactions and their targets. First described in 2006, Drug- Bank has continued to evolve over the past 12 years in response to marked improvements to web stds. and changing needs for drug research and development. This year's update, DrugBank 5.0, represents the most significant upgrade to the database in more than 10 years. In many cases, existing data content has grown by 100% or more over the last update. For instance, the total no. of investigational drugs in the database has grown by almost 300%, the no. of drug-drug interactions has grown by nearly 600% and the no. of SNP-assocd. drug effects has grown more than 3000%. Significant improvements have been made to the quantity, quality and consistency of drug indications, drug binding data as well as drug-drug and drug-food interactions. A great deal of brand new data have also been added to DrugBank 5.0. This includes information on the influence of hundreds of drugs on metabolite levels (pharmacometabolomics), gene expression levels (pharmacotranscriptomics) and protein expression levels (pharmacoproteomics). New data have also been added on the status of hundreds of newdrug clin. trials and existing drug repurposing trials. Many other important improvements in the content, interface and performance of the DrugBank website have been made and these should greatly enhance its ease of use, utility and potential applications in many areas of pharmacol. research, pharmaceutical science and drug education.
  42. 42
    Pracht, P.; Bohle, F.; Grimme, S. Automated Exploration of the low-energy Chemical Space with fast Quantum Chemical Methods. Phys. Chem. Chem. Phys. 2020, 22, 7169– 7192,  DOI: 10.1039/C9CP06869D
    Google Scholar
    42
    Automated exploration of the low-energy chemical space with fast quantum chemical methods
    Pracht, Philipp; Bohle, Fabian; Grimme, Stefan
    Physical Chemistry Chemical Physics (2020), 22 (14), 7169-7192CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
    We propose and discuss an efficient scheme for the in silico sampling for parts of the mol. chem. space by semiempirical tight-binding methods combined with a meta-dynamics driven search algorithm. The focus of this work is set on the generation of proper thermodn. ensembles at a quantum chem. level for conformers, but similar procedures for protonation states, tautomerism and non-covalent complex geometries are also discussed. The conformational ensembles consisting of all significantly populated min. energy structures normally form the basis of further, mostly DFT computational work, such as the calcn. of spectra or macroscopic properties. By using basic quantum chem. methods, electronic effects or possible bond breaking/formation are accounted for and a very reasonable initial energetic ranking of the candidate structures is obtained. Due to the huge computational speedup gained by the fast low-cost quantum chem. methods, overall short computation times even for systems with hundreds of atoms (typically drug-sized mols.) are achieved. Furthermore, specialized applications, such as sampling with implicit solvation models or constrained conformational sampling for transition-states, metal-, surface-, or noncovalently bound complexes are discussed, opening many possible applications in modern computational chem. and drug discovery. The procedures have been implemented in a freely available computer code called CREST, that makes use of the fast and reliable GFNn-xTB methods.
  43. 43
    Grimme, S. Exploration of Chemical Compound, Conformer, and Reaction Space with Meta-Dynamics Simulations Based on Tight-Binding Quantum Chemical Calculations. J. Chem. Theory Comput. 2019, 15, 2847– 2862,  DOI: 10.1021/acs.jctc.9b00143
    Google Scholar
    43
    Exploration of Chemical Compound, Conformer, and Reaction Space with Meta-Dynamics Simulations Based on Tight-Binding Quantum Chemical Calculations
    Grimme, Stefan
    Journal of Chemical Theory and Computation (2019), 15 (5), 2847-2862CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
    The semiempirical tight-binding based quantum chem. method GFN2-xTB is used in the framework of meta-dynamics (MTD) to globally explore chem. compd., conformer, and reaction space. The biasing potential given as a sum of Gaussian functions is expressed with the root-mean-square-deviation (RMSD) in Cartesian space as a metric for the collective variables. This choice makes the approach robust and generally applicable to three common problems (i.e., conformer search, chem. reaction space exploration in a virtual nanoreactor, and for guessing reaction paths). Because of the inherent locality of the at. RMSD, functional group or fragment selective treatments are possible facilitating the investigation of catalytic processes where, for example, only the substrate is thermally activated. Due to the approx. character of the GFN2-xTB method, the resulting structure ensembles require further refinement with more sophisticated, for example, d. functional or wave function theory methods. However, the approach is extremely efficient running routinely on common laptop computers in minutes to hours of computation time even for realistically sized mols. with a few hundred atoms. Furthermore, the underlying potential energy surface for mols. contg. almost all elements (Z = 1-86) is globally consistent including the covalent dissocn. process and electronically complicated situations in, for example, transition metal systems. As examples, thermal decompn., ethyne oligomerization, the oxidn. of hydrocarbons (by oxygen and a P 450 enzyme model), a Miller-Urey model system, a thermally forbidden dimerization, and a multistep intramol. cyclization reaction are shown. For typical conformational search problems of org. drug mols., the new MTD(RMSD) algorithm yields lower energy structures and more complete conformer ensembles at reduced computational effort compared with its already well performing predecessor.
  44. 44
    Chan, L.; Morris, G. M.; Hutchison, G. R. Understanding Conformational Entropy in Small Molecules. J. Chem. Theory Comput. 2021, 17, 2099– 2106,  DOI: 10.1021/acs.jctc.0c01213
    Google Scholar
    44
    Understanding Conformational Entropy in Small Molecules
    Chan, Lucian; Morris, Garrett M.; Hutchison, Geoffrey R.
    Journal of Chemical Theory and Computation (2021), 17 (4), 2099-2106CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
    The calcn. of the entropy of flexible mols. can be challenging, since the no. of possible conformers can grow exponentially with mol. size and many low-energy conformers may be thermally accessible. Different methods have been proposed to approx. the contribution of conformational entropy to the mol. std. entropy, including performing thermochem. calcns. with all possible stable conformations and developing empirical corrections from exptl. data. We have performed conformer sampling on over 120,000 small mols. generating some 12 million conformers, to develop models to predict conformational entropy across a wide range of mols. Using insight into the nature of conformational disorder, our cross-validated phys. motivated statistical model gives a mean abs. error of ∼ 4.8 J/mol·K or under 0.4 kcal/mol at 300 K. Beyond predicting mol. entropies and free energies, the model implies a high degree of correlation between torsions in most mols., often assumed to be independent. While individual dihedral rotations may have low energetic barriers, the shape and chem. functionality of most mols. necessarily correlate their torsional degrees of freedom and hence restrict the no. of low-energy conformations immensely. Our simple models capture these correlations and advance our understanding of small mol. conformational entropy.
  45. 45
    Landrum, G. RDKit: Open-Source Cheminformatics. Available at http://www.rdkit.org, 2020; http://www.rdkit.org (accesses Oct 1, 2022).
    Google Scholar
    There is no corresponding record for this reference.
  46. 46
    https://smarts.plus/.
    Google Scholar
    There is no corresponding record for this reference.
  47. 47
    Schomburg, K.; Ehrlich, H.-C.; Stierand, K.; Rarey, M. From Structure Diagrams to Visual Chemical Patterns. J. Chem. Inf. Model. 2010, 50, 1529– 1535,  DOI: 10.1021/ci100209a
    Google Scholar
    47
    From Structure Diagrams to Visual Chemical Patterns
    Schomburg, Karen; Ehrlich, Hans-Christian; Stierand, Katrin; Rarey, Matthias
    Journal of Chemical Information and Modeling (2010), 50 (9), 1529-1535CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The intuitive way of chemists to communicate mols. is via two-dimensional structure diagrams. The straightforward visual representations are mostly preferred to the often complicated systematic chem. names. For chem. patterns, however, no comparable visualization stds. have evolved so far. Chem. patterns denoting descriptions of chem. features are needed whenever a set of mols. is filtered for certain properties. The currently available representations are constrained to linear mol. pattern languages which are hardly human readable and therefore keep chemists without computational background from systematically formulating patterns. Therefore, we introduce a new visualization concept for chem. patterns. The common std. concept of structure diagrams is extended to account for property descriptions and logic combinations of chem. features in patterns. As a first application of the new concept, we developed the SMARTSviewer, a tool that converts chem. patterns encoded in SMARTS strings to a visual representation. The graphic pattern depiction provides an overview of the specified chem. features, variations, and similarities without needing to decode the often cryptic linear expressions. Taking recent chem. publications from various fields, we demonstrate the wide application range of a graphical chem. pattern language.
  48. 48
    Neese, F.; Wennmohs, F.; Becker, U.; Riplinger, C. The ORCA quantum chemistry program package. J. Chem. Phys. 2020, 152, 224108,  DOI: 10.1063/5.0004608
    Google Scholar
    48
    The ORCA quantum chemistry program package
    Neese, Frank; Wennmohs, Frank; Becker, Ute; Riplinger, Christoph
    Journal of Chemical Physics (2020), 152 (22), 224108CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
    In this contribution to the special software-centered issue, the ORCA program package is described. We start with a short historical perspective of how the project began and go on to discuss its current feature set. ORCA has grown into a rather comprehensive general-purpose package for theor. research in all areas of chem. and many neighboring disciplines such as materials sciences and biochem. ORCA features d. functional theory, a range of wavefunction based correlation methods, semi-empirical methods, and even force-field methods. A range of solvation and embedding models is featured as well as a complete intrinsic to ORCA quantum mechanics/mol. mechanics engine. A specialty of ORCA always has been a focus on transition metals and spectroscopy as well as a focus on applicability of the implemented methods to "real-life" chem. applications involving systems with a few hundred atoms. In addn. to being efficient, user friendly, and, to the largest extent possible, platform independent, ORCA features a no. of methods that are either unique to ORCA or have been first implemented in the course of the ORCA development. Next to a range of spectroscopic and magnetic properties, the linear- or low-order single- and multi-ref. local correlation methods based on pair natural orbitals (domain based local pair natural orbital methods) should be mentioned here. Consequently, ORCA is a widely used program in various areas of chem. and spectroscopy with a current user base of over 22 000 registered users in academic research and in industry. (c) 2020 American Institute of Physics.
  49. 49
    Lin, J. B.; Jin, Y.; Lopez, S. A.; Druckerman, N.; Wheeler, S. E.; Houk, K. N. Torsional Barriers to Rotation and Planarization in Heterocyclic Oligomers of Value in Organic Electronics. J. Chem. Theory Comput. 2017, 13, 5624– 5638,  DOI: 10.1021/acs.jctc.7b00709
    Google Scholar
    49
    Torsional Barriers to Rotation and Planarization in Heterocyclic Oligomers of Value in Organic Electronics
    Lin, Janice B.; Jin, Yu; Lopez, Steven A.; Druckerman, Nathaniel; Wheeler, Steven E.; Houk, K. N.
    Journal of Chemical Theory and Computation (2017), 13 (11), 5624-5638CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
    In order to understand the conformational behavior of org. components in org. electronic devices, we have computed the torsional potentials for a library of thiophene-based heterodimers. The accuracy and efficiencies of computational methods for these org. materials were benchmarked for 11 common d. functionals with three Pople basis sets against a Focal Point Anal. (FPA) on a model oligothiophene 2,5-bis(3-tetradecylthiophen-2-yl)thieno[3,2-b]-thiophene (BTTT) system. This study establishes a set of general trends in regards to conformational preferences, as well as planarization and rotational barriers for a library comprised of common fragments found in org. materials. These gas phase structures are compared to exptl. crystal structures to det. the effect of crystal packing on geometry. Finally, we analyze the structure of hole-transporting material DERDTS-TBDT and design a new oligomer likely to be planar in the solid state.
  50. 50
    Perkins, M. A.; Cline, L. M.; Tschumper, G. S. Torsional Profiles of Thiophene and Furan Oligomers: Probing the Effects of Heterogeneity and Chain Length. J. Phys. Chem. A 2021, 125, 6228– 6237,  DOI: 10.1021/acs.jpca.1c04714
    Google Scholar
    50
    Torsional Profiles of Thiophene and Furan Oligomers: Probing the Effects of Heterogeneity and Chain Length
    Perkins, Morgan A.; Cline, Laura M.; Tschumper, Gregory S.
    Journal of Physical Chemistry A (2021), 125 (28), 6228-6237CODEN: JPCAFH; ISSN:1089-5639. (American Chemical Society)
    A systematic anal. of the torsional profiles of 55 unique oligomers composed of two to four thiophene and/or furan rings (n = 2 to 4) has been conducted using three d. functional theory (DFT) methods along with MP2 and three different coupled-cluster methods. Two planar or quasi-planar min. were identified for each n = 2 oligomer system. In every case, the torsional angle (τ) between the heteroatoms about the carbon-carbon bond connecting the two rings is at or near 180° for the global min. and 0° for the local min., referred to as anti and syn conformations, resp. These oligomers have rotational barrier heights ranging from ca. 2 kcal mol-1 for 2,2'-bithiophene to 4 kcal mol-1 for 2,2'-bifuran, based on electronic energies computed near the CCSD(T) complete basis set (CBS) limit. The corresponding rotational barrier for the heterogeneous 2-(2-thienyl)furan counterpart falls approx. halfway between those values. The energy differences between the min. are approx. 2 and 0.4 kcal mol-1 for the homogeneous 2,2'-bifuran and 2,2'-bithiophene, resp., whereas the energy difference between the planar local and global min. (at τ = 0 and 180°, resp.) is only 0.3 kcal mol-1 for 2-(2-thienyl)furan. Extending these three oligomers by adding one or two addnl. thiophene and/or furan rings resulted in only minor changes to the torsional profiles when rotating around the same carbon-carbon bond as the two-ring profiles. Relative energy differences between the syn and anti conformations were changed by no more than 0.4 kcal mol-1 for the corresponding n = 3 and 4 oligomers, while the rotational barrier height increased by no more than 0.8 kcal mol-1.
  51. 51
    Johansson, M. P.; Olsen, J. Torsional Barriers and Equilibrium Angle of Biphenyl: Reconciling Theory with Experiment. J. Chem. Theory Comput. 2008, 4, 1460– 1471,  DOI: 10.1021/ct800182e
    Google Scholar
    51
    Torsional Barriers and Equilibrium Angle of Biphenyl: Reconciling Theory with Experiment
    Johansson, Mikael P.; Olsen, Jeppe
    Journal of Chemical Theory and Computation (2008), 4 (9), 1460-1471CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
    The barriers of internal rotation of the two Ph groups in biphenyl are investigated using a combination of coupled cluster and d. functional theory. The exptl. barriers are for the first time accurately reproduced; our best ests. of the barriers are 8.0 and 8.3 kJ/mol around the planar and perpendicular conformations, resp. The use of flexible basis sets of at least augmented quadruple-ζ quality is shown to be a crucial prerequisite. Further, to finally reconcile theory with expt., extrapolations of both the basis set toward the basis set limit and electron correlation toward the full configuration-interaction limit are necessary. The min. of the torsional angle is significantly increased by free energy corrections, which are needed to reach an agreement with expt. The d. functional B3LYP approach is found to perform well compared with the highest level ab initio results.
  52. 52
    Nam, S.; Cho, E.; Sim, E.; Burke, K. Explaining and Fixing DFT Failures for Torsional Barriers. J. Phys. Chem. Lett. 2021, 12, 2796– 2804,  DOI: 10.1021/acs.jpclett.1c00426
    Google Scholar
    52
    Explaining and Fixing DFT Failures for Torsional Barriers
    Nam, Seungsoo; Cho, Eunbyol; Sim, Eunji; Burke, Kieron
    Journal of Physical Chemistry Letters (2021), 12 (11), 2796-2804CODEN: JPCLCD; ISSN:1948-7185. (American Chemical Society)
    Most torsional barriers are predicted with high accuracies (about 1 kJ/mol) by std. semilocal functionals, but a small subset was found to have much larger errors. We created a database of almost 300 carbon-carbon torsional barriers, including 12 poorly behaved barriers, that stem from the Y=C-X group, where Y is O or S and X is a halide. Functionals with enhanced exchange mixing (about 50%) worked well for all barriers. We found that poor actors have delocalization errors caused by hyperconjugation. These problematic calcns. are d.-sensitive (i.e., DFT predictions change noticeably with the d.), and using HF densities (HF-DFT) fixes these issues. For example, conventional B3LYP performs as accurately as exchange-enhanced functionals if the HF d. is used. For long-chain conjugated mols., HF-DFT can be much better than exchange-enhanced functionals. We suggest that HF-PBE0 has the best overall performance.
  53. 53
    Jackson, N. E.; Savoie, B. M.; Kohlstedt, K. L.; Olvera de la Cruz, M.; Schatz, G. C.; Chen, L. X.; Ratner, M. A. Controlling Conformations of Conjugated Polymers and Small Molecules: The Role of Nonbonding Interactions. J. Am. Chem. Soc. 2013, 135, 10475– 10483,  DOI: 10.1021/ja403667s
    Google Scholar
    53
    Controlling Conformations of Conjugated Polymers and Small Molecules: The Role of Nonbonding Interactions
    Jackson, Nicholas E.; Savoie, Brett M.; Kohlstedt, Kevin L.; Olvera de la Cruz, Monica; Schatz, George C.; Chen, Lin X.; Ratner, Mark A.
    Journal of the American Chemical Society (2013), 135 (28), 10475-10483CODEN: JACSAT; ISSN:0002-7863. (American Chemical Society)
    The chem. variety present in the org. electronics literature has motivated us to investigate potential nonbonding interactions often incorporated into conformational "locking" schemes. We examine a variety of potential interactions, including oxygen-sulfur, nitrogen-sulfur, and fluorine-sulfur, using accurate quantum-chem. wave function methods and noncovalent interaction (NCI) anal. on a selection of high-performing conjugated polymers and small mols. found in the literature. In addn., we evaluate a set of nonbonding interactions occurring between various heterocyclic and pendant atoms taken from a group of representative π-conjugated mols. Together with our survey and set of interactions, it is detd. that while many nonbonding interactions possess weak binding capabilities, nontraditional hydrogen-bonding interactions, oxygen-hydrogen (CH···O) and nitrogen-hydrogen (CH···N), are alone in inducing conformational control and enhanced planarity along a polymer or small mol. backbone at room temp.
  54. 54
    Greenwell, C.; Beran, G. J. O. Inaccurate Conformational Energies Still Hinder Crystal Structure Prediction in Flexible Organic Molecules. Cryst. Growth Des. 2020, 20, 4875– 4881,  DOI: 10.1021/acs.cgd.0c00676
    Google Scholar
    54
    Inaccurate Conformational Energies Still Hinder Crystal Structure Prediction in Flexible Organic Molecules
    Greenwell, Chandler; Beran, Gregory J. O.
    Crystal Growth & Design (2020), 20 (8), 4875-4881CODEN: CGDEFU; ISSN:1528-7483. (American Chemical Society)
    Crystal structure prediction driven by d. functional theory has become an increasingly useful tool for the pharmaceutical industry and others interested in understanding and controlling org. mol. crystal packing. However, delocalization error in widely used d. functionals leads to problematic conformational energies that can cause incorrect predictions of polymorph stabilities. In 5 examples ranging from small mols. to the polymorphically challenging pharmaceuticals axitinib and galunisertib, inexpensively correcting the intramol. conformational energies with higher-level electronic structure methods leads to polymorph stability predictions that agree far better with expt. This approach also provides a valuable diagnostic for when skepticism about predicted polymorph stabilities is warranted. Commonly used d. functionals have difficulty ranking certain types of conformational polymorph structures correctly. Correcting the intramol. conformational energies with higher-level quantum chem. methods can improve the accuracy of crystal structure prediction stability rankings considerably.
  55. 55
    Smith, J. S.; Zubatyuk, R.; Nebgen, B.; Lubbers, N.; Barros, K.; Roitberg, A. E.; Isayev, O.; Tretiak, S. The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules. Sci. Data 2020, 7, 134,  DOI: 10.1038/s41597-020-0473-z
    Google Scholar
    55
    The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules
    Smith, Justin S.; Zubatyuk, Roman; Nebgen, Benjamin; Lubbers, Nicholas; Barros, Kipton; Roitberg, Adrian E.; Isayev, Olexandr; Tretiak, Sergei
    Scientific Data (2020), 7 (1), 134CODEN: SDCABS; ISSN:2052-4463. (Nature Research)
    Abstr.: Maximum diversification of data is a central theme in building generalized and accurate machine learning (ML) models. In chem., ML has been used to develop models for predicting mol. properties, for example quantum mechanics (QM) calcd. potential energy surfaces and at. charge models. The ANI-1x and ANI-1ccx ML-based general-purpose potentials for org. mols. were developed through active learning; an automated data diversification process. Here, we describe the ANI-1x and ANI-1ccx data sets. To demonstrate data diversity, we visualize it with a dimensionality redn. scheme, and contrast against existing data sets. The ANI-1x data set contains multiple QM properties from 5 M d. functional theory calcns., while the ANI-1ccx data set contains 500 k data points obtained with an accurate CCSD(T)/CBS extrapolation. Approx. 14 million CPU core-hours were expended to generate this data. Multiple QM calcd. properties for the chem. elements C, H, N, and O are provided: energies, at. forces, multipole moments, at. charges, etc. We provide this data to the community to aid research and development of ML models for chem.
  56. 56
    Axelrod, S.; Gomez-Bombarelli, R. GEOM, energy-annotated molecular conformations for property prediction and molecular generation. Sci. Data 2022, 9, 185,  DOI: 10.1038/s41597-022-01288-4
    Google Scholar
    56
    GEOM, energy-annotated molecular conformations for property prediction and molecular generation
    Axelrod, Simon; Gomez-Bombarelli, Rafael
    Scientific Data (2022), 9 (1), 185CODEN: SDCABS; ISSN:2052-4463. (Nature Portfolio)
    Machine learning (ML) outperforms traditional approaches in many mol. design tasks. ML models usually predict mol. properties from a 2D chem. graph or a single 3D structure, but neither of these representations accounts for the ensemble of 3D conformers that are accessible to a mol. Property prediction could be improved by using conformer ensembles as input, but there is no large-scale dataset that contains graphs annotated with accurate conformers and exptl. data. Here we use advanced sampling and semi-empirical d. functional theory (DFT) to generate 37 million mol. conformations for over 450,000 mols. The Geometric Ensemble Of Mols. (GEOM) dataset contains conformers for 133,000 species from QM9, and 317,000 species with exptl. data related to biophysics, physiol., and phys. chem. Ensembles of 1,511 species with BACE-1 inhibition data are also labeled with high-quality DFT free energies in an implicit water solvent, and 534 ensembles are further optimized with DFT. GEOM will assist in the development of models that predict properties from conformer ensembles, and generative models that sample 3D conformations.
  57. 57
    Isert, C.; Atz, K.; Jiménez-Luna, J.; Schneider, G. QMugs, quantum mechanical properties of drug-like molecules. Sci. Data 2022, 9, 273,  DOI: 10.1038/s41597-022-01390-7
    Google Scholar
    57
    QMugs, quantum mechanical properties of drug-like molecules
    Isert, Clemens; Atz, Kenneth; Jimenez-Luna, Jose; Schneider, Gisbert
    Scientific Data (2022), 9 (1), 273CODEN: SDCABS; ISSN:2052-4463. (Nature Portfolio)
    Machine learning approaches in drug discovery, as well as in other areas of the chem. sciences, benefit from curated datasets of phys. mol. properties. However, there currently is a lack of data collections featuring large bioactive mols. alongside first-principle quantum chem. information. The open-access QMugs (Quantum-Mech. Properties of Drug-like Mols.) dataset fills this void. The QMugs collection comprises quantum mech. properties of more than 665 k biol. and pharmacol. relevant mols. extd. from the ChEMBL database, totaling ∼2 M conformers. QMugs contains optimized mol. geometries and thermodn. data obtained via the semi-empirical method GFN2-xTB. Atomic and mol. properties are provided on both the GFN2-xTB and on the d.-functional levels of theory (DFT, ωB97X-D/def2-SVP). QMugs features mols. of significantly larger size than previously-reported collections and comprises their resp. quantum mech. wave functions, including DFT d. and orbital matrixes. This dataset is intended to facilitate the development of models that learn from mol. data on different levels of theory while also providing insight into the corresponding relationships between mol. structure and biol. activity.
  58. 58
    Eastman, P.; Behara, P. K.; Dotson, D. L.; Galvelis, R.; Herr, J. E.; Horton, J. T.; Mao, Y.; Chodera, J. D.; Pritchard, B. P.; Wang, Y.; Fabritiis, G. D.; Markland, T. E. SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials. Sci. Data 2023, 10, 11,  DOI: 10.1038/s41597-022-01882-6
    Google Scholar
    58
    SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials
    Eastman, Peter; Behara, Pavan Kumar; Dotson, David L.; Galvelis, Raimondas; Herr, John E.; Horton, Josh T.; Mao, Yuezhi; Chodera, John D.; Pritchard, Benjamin P.; Wang, Yuanqing; De Fabritiis, Gianni; Markland, Thomas E.
    Scientific Data (2023), 10 (1), 11CODEN: SDCABS; ISSN:2052-4463. (Nature Portfolio)
    Machine learning potentials are an important tool for mol. simulation, but their development is held back by a shortage of high quality datasets to train them on. We describe the SPICE dataset, a new quantum chem. dataset for training potentials relevant to simulating drug-like small mols. interacting with proteins. It contains over 1.1 million conformations for a diverse set of small mols., dimers, dipeptides, and solvated amino acids. It includes 15 elements, charged and uncharged mols., and a wide range of covalent and non-covalent interactions. It provides both forces and energies calcd. at the ωB97M-D3(BJ)/def2-TZVPPD level of theory, along with other useful quantities such as multipole moments and bond orders. We train a set of machine learning potentials on it and demonstrate that they can achieve chem. accuracy across a broad region of chem. space. It can serve as a valuable resource for the creation of transferable, ready to use potential functions for use in mol. simulations.
  59. 59
    McNutt, A. T.; Bisiriyu, F.; Song, S.; Vyas, A.; Hutchison, G. R.; Koes, D. R. Conformer Generation for Structure-Based Drug Design: How Many and How Good?. J. Chem. Inf. Model. 2023, 63, 6598– 6607,  DOI: 10.1021/acs.jcim.3c01245
    Google Scholar
    59
    Conformer Generation for Structure-Based Drug Design: How Many and How Good?
    McNutt, Andrew T.; Bisiriyu, Fatimah; Song, Sophia; Vyas, Ananya; Hutchison, Geoffrey R.; Koes, David Ryan
    Journal of Chemical Information and Modeling (2023), 63 (21), 6598-6607CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    Conformer generation, the assignment of realistic 3D coordinates to a small mol., is fundamental to structure-based drug design. Conformational ensembles are required for rigid-body matching algorithms, such as shape-based or pharmacophore approaches, and even methods that treat the ligand flexibly, such as docking, are dependent on the quality of the provided conformations due to not sampling all degrees of freedom (e.g., only sampling torsions). Here, we empirically elucidate some general principles about the size, diversity, and quality of the conformational ensembles needed to get the best performance in common structure-based drug discovery tasks. In many cases, our findings may parallel "common knowledge" well-known to practitioners of the field. Nonetheless, we feel that it is valuable to quantify these conformational effects while reproducing and expanding upon previous studies. Specifically, we investigate the performance of a state-of-the-art generative deep learning approach vs. a more classical geometry-based approach, the effect of energy minimization as a postprocessing step, the effect of ensemble size (max. no. of conformers), and construction (filtering by root-mean-square deviation for diversity) and how these choices influence the ability to recapitulate bioactive conformations and perform pharmacophore screening and mol. docking.
  60. 60
    Foloppe, N.; Chen, I.-J. Energy windows for computed compound conformers: covering artefacts or truly large reorganization energies?. Future Med. Chem. 2019, 11, 97– 118,  DOI: 10.4155/fmc-2018-0400
    Google Scholar
    60
    Energy windows for computed compound conformers: covering artefacts or truly large reorganization energies?
    Foloppe, Nicolas; Chen, I-Jen
    Future Medicinal Chemistry (2019), 11 (2), 97-118CODEN: FMCUA7; ISSN:1756-8919. (Future Science Ltd.)
    The generation of 3D conformers of small mols. underpins most computational drug discovery. Thus, the conformer quality is crit. and depends on their energetics. A key parameter is the empirical conformational energy window (ΔEw), since only conformers within ΔEw are retained. However, ΔEw values in use appear unrealistically large. We analyze the factors pertaining to the conformer energetics and ΔEw. We argue that more attention must be focused on the problem of collapsed low-energy conformers. That is due to artificial intramol. stabilization and occurs even with continuum solvation. Consequently, the conformational energy of extended bioactive structures is artifactually increased, which inflates ΔEw. Thus, this Perspective highlights the issues arising from low-energy conformers and suggests improvements via empirical or physics-based strategies. Graphical abstr. :.
  61. 61
    Rai, B. K.; Sresht, V.; Yang, Q.; Unwalla, R.; Tu, M.; Mathiowetz, A. M.; Bakken, G. A. Comprehensive Assessment of Torsional Strain in Crystal Structures of Small Molecules and Protein–Ligand Complexes using ab Initio Calculations. J. Chem. Inf. Model. 2019, 59, 4195– 4208,  DOI: 10.1021/acs.jcim.9b00373
    Google Scholar
    61
    Comprehensive Assessment of Torsional Strain in Crystal Structures of Small Molecules and Protein-Ligand Complexes using ab Initio Calculations
    Rai, Brajesh K.; Sresht, Vishnu; Yang, Qingyi; Unwalla, Ray; Tu, Meihua; Mathiowetz, Alan M.; Bakken, Gregory A.
    Journal of Chemical Information and Modeling (2019), 59 (10), 4195-4208CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The energetics of rotation around single bonds (torsions) is a key determinant of the 3D shape that drug-like mols. adopt in soln., the solid state, and in different biol. environments, which in turn defines their unique phys. and pharmacol. properties. Therefore, accurate characterization of torsion angle preference and energetics is essential for the success of computational drug discovery and design. Here, the authors analyze torsional strain in crystal structures of drug-like mols. in CSD and bioactive ligand conformations in PDB, expressing the total strain energy as a sum of strain energy from constituent rotatable bonds. The authors utilized Cloud computing to generate torsion scan profiles of a very large collection of chem. diverse neutral fragments at DFT(B3LYP)/6-31G*//6-31G** or DFT(B3LYP)/6-31+G*//6-31+G** (for sulfur-contg. mol.). With the data generated from these ab initio calcns., the authors performed rigorous anal. of strain due to deviation of obsd. torsion angles relative to their ideal gas-phase geometries. Contrary to the previous studies based on mol. mechanics, the authors find that in the cryst.-state mols. generally adopt low-strain conformations, with median per-torsion strain energy in CSD and PDB under 1/10th and 1/3rd of a kcal/mol, resp. However, for a small fraction (<5%) of motifs, external effects such as steric hindrance and hydrogen bonds result in strain penalty exceeding 2.5 kcal/mol. The authors find that due to poor quality of PDB structures in general, bioactive structures tend to have higher torsional strain compared to small mol. crystal conformations. However, in the absence of structural fitting artifacts in PDB structures, protein-induced strain in bioactive conformations is quant. similar to those due to the packing forces in small mol. crystal structures. This anal. allows us to establish strain energy thresholds to help to identify biol. relevant conformers in a given ensemble. The work presented here is the most comprehensive study to date that demonstrates the utility and feasibility of gas-phase QM calcns. to study conformational preference and energetics of drug-size mols. Potential applications of this study in computational lead discovery and structure-based design are discussed.
  62. 62
    Taylor, R.; Wood, P. A. A Million Crystal Structures: The Whole Is Greater than the Sum of Its Parts. Chem. Rev. 2019, 119, 9427– 9477,  DOI: 10.1021/acs.chemrev.9b00155
    Google Scholar
    62
    A Million Crystal Structures: The Whole Is Greater than the Sum of Its Parts
    Taylor, Robin; Wood, Peter A.
    Chemical Reviews (Washington, DC, United States) (2019), 119 (16), 9427-9477CODEN: CHREAY; ISSN:0009-2665. (American Chemical Society)
    A review. The founding in 1965 of what is now called the Cambridge Structural Database (CSD) has reaped dividends in numerous and diverse areas of chem. research. Each of the million or so crystal structures in the database was solved for its own particular reason, but collected together, the structures can be reused to address a multitude of new problems. In this Review, which is focused mainly on the last 10 years, we chronicle the contribution of the CSD to research into mol. geometries, mol. interactions, and mol. assemblies and demonstrate its value in the design of biol. active mols. and the solid forms in which they are delivered. Its potential in other com. relevant areas is described, including gas storage and delivery, thin films, and (opto)electronics. The CSD also aids the soln. of new crystal structures. Because no scientific instrument is without shortcomings, the limitations of CSD research are assessed. We emphasize the importance of maintaining database quality: notwithstanding the arrival of big data and machine learning, it remains perilous to ignore the principle of garbage in, garbage out. Finally, we explain why the CSD must evolve with the world around it to ensure it remains fit for purpose in the years ahead.
  63. 63
    Liebeschuetz, J. W. The Good, the Bad, and the Twisted Revisited: An Analysis of Ligand Geometry in Highly Resolved Protein–Ligand X-ray Structures. J. Med. Chem. 2021, 64, 7533– 7543,  DOI: 10.1021/acs.jmedchem.1c00228
    Google Scholar
    63
    The Good, the Bad, and the Twisted Revisited: An Analysis of Ligand Geometry in Highly Resolved Protein-Ligand X-ray Structures
    Liebeschuetz, John W.
    Journal of Medicinal Chemistry (2021), 64 (11), 7533-7543CODEN: JMCMAR; ISSN:0022-2623. (American Chemical Society)
    An anal. of the rotatable bond geometry of drug-like ligand models is reported for high-resoln. (<1.1 Å) crystallog. protein-ligand complexes. In cases where the ligand fit to the electron d. is very good, unusual torsional geometry is rare and, most often, though not exclusively, assocd. with strong polar, metal, or covalent ligand-protein interactions. It is rarely assocd. with a torsional strain of greater than 2 kcal mol-1 by calcn. An unusual torsional geometry is more prevalent where the fit to electron d. is not perfect. Multiple low-strain conformer bindings were obsd. in 21% of the set and, it is suggested, may also lie behind many of the 35% of single-occupancy cases, where a poor fit to the e-d. was found. It is concluded that multiple conformer ligand binding is an under-recognized phenomenon in structure-based drug design and that there is a need for more robust crystallog. refinement methods to better handle such cases.
  64. 64
    Tong, J.; Zhao, S. Large-Scale Analysis of Bioactive Ligand Conformational Strain Energy by Ab Initio Calculation. J. Chem. Inf. Model. 2021, 61, 1180– 1192,  DOI: 10.1021/acs.jcim.0c01197
    Google Scholar
    64
    Large-Scale Analysis of Bioactive Ligand Conformational Strain Energy by Ab Initio Calculation
    Tong, Jiahui; Zhao, Suwen
    Journal of Chemical Information and Modeling (2021), 61 (3), 1180-1192CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    Ligand conformational strain energy (LCSE) plays an important role in virtual screening and lead optimization. While various studies have provided insights into LCSE for small-mol. ligands in the Protein Data Bank (PDB), conclusions are inconsistent mainly due to small datasets, poor quality control of crystal structures, and mol. mechanics (MM) or low-level quantum mechanics (QM) calcns. Here, we built a high-quality dataset (LigBoundConf) of 8145 ligand-bound conformations from PDB crystal structures and calcd. LCSE at the M062X-D3/ma-TZVPP (SMD)//M062X-D3/def2-SVP(SMD) level for each case in the dataset. The mean/median LCSE is 4.6/3.7 kcal/mol for 6672 successfully calcd. cases, which is significantly lower than the ests. based on mol. mechanics in many previous analyses. Esp., when removing ligands with nonarom. ring(s) that are prone to have large LCSEs due to electron d. overfitting, the mean/median LCSE was reduced to 3.3/2.5 kcal/mol. We further reveal that LCSE is correlated with several ligand properties, including formal at. charge, mol. wt., no. of rotatable bonds, and no. of hydrogen-bond donors and acceptors. In addn., our results show that although summation of torsion strains is a good approxn. of LCSE for most cases, for a small fraction (about 6%) of our dataset, it underestimates LCSEs if ligands could form nonlocal intramol. interactions in the unbound state. Taken together, our work provides a comprehensive profile of LCSE for ligands in PDB, which could help ligand conformation generation, ligand docking pose evaluation, and lead optimization.
  65. 65
    Chan, L.; Hutchison, G. R.; Morris, G. M. Understanding Ring Puckering in Small Molecules and Cyclic Peptides. J. Chem. Inf. Model. 2021, 61, 743– 755,  DOI: 10.1021/acs.jcim.0c01144
    Google Scholar
    65
    Understanding Ring Puckering in Small Molecules and Cyclic Peptides
    Chan, Lucian; Hutchison, Geoffrey R.; Morris, Garrett M.
    Journal of Chemical Information and Modeling (2021), 61 (2), 743-755CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
    The geometry of a mol. plays a significant role in detg. its phys. and chem. properties. Despite its importance, there are relatively few studies on ring puckering and conformations, often focused on small cycloalkanes, 5- and 6-membered carbohydrate rings, and specific macrocycle families. We lack a general understanding of the puckering preferences of medium-sized rings and macrocycles. To address this, we provide an extensive conformational anal. of a diverse set of rings. We used Cremer-Pople puckering coordinates to study the trends of the ring conformation across a set of 140 000 diverse small mols., including small rings, macrocycles, and cyclic peptides. By standardizing using key atoms, we show that the ring conformations can be classified into relatively few conformational clusters, based on their canonical forms. The no. of such canonical clusters increases slowly with ring size. Ring puckering motions, esp. pseudo-rotations, are generally restricted and differ between clusters. More importantly, we propose models to map puckering preferences to torsion space, which allows us to understand the inter-related changes in torsion angles during pseudo-rotation and other puckering motions. Beyond ring puckers, our models also explain the change in substituent orientation upon puckering. We also present a novel knowledge-based sampling method using the puckering preferences and coupled substituent motion to generate ring conformations efficiently. In summary, this work provides an improved understanding of general ring puckering preferences, which will in turn accelerate the identification of low-energy ring conformations for applications from polymeric materials to drug binding.
  66. 66
    Lemm, D.; von Rudorff, G. F.; von Lilienfeld, O. A. Machine learning based energy-free structure predictions of molecules, transition states, and solids. Nat. Commun. 2021, 12, 4468,  DOI: 10.1038/s41467-021-24525-7
    Google Scholar
    66
    Machine learning based energy-free structure predictions of molecules, transition states, and solids
    Lemm, Dominik; von Rudorff, Guido Falk; von Lilienfeld, O. Anatole
    Nature Communications (2021), 12 (1), 4468CODEN: NCAOBW; ISSN:2041-1723. (Nature Research)
    The computational prediction of atomistic structure is a long-standing problem in physics, chem., materials, and biol. Conventionally, force-fields or ab initio methods det. structure through energy minimization, which is either approx. or computationally demanding. This accuracy/cost trade-off prohibits the generation of synthetic big data sets accounting for chem. space with atomistic detail. Exploiting implicit correlations among relaxed structures in training data sets, our machine learning model Graph-To-Structure (G2S) generalizes across compd. space in order to infer interat. distances for out-of-sample compds., effectively enabling the direct reconstruction of coordinates, and thereby bypassing the conventional energy optimization task. The numerical evidence collected includes 3D coordinate predictions for org. mols., transition states, and cryst. solids. G2S improves systematically with training set size, reaching mean abs. interat. distance prediction errors of less than 0.2 Å for less than eight thousand training structures - on par or better than conventional structure generators. Applicability tests of G2S include successful predictions for systems which typically require manual intervention, improved initial guesses for subsequent conventional ab initio based relaxation, and input generation for subsequent use of structure based quantum machine learning models.
  67. 67
    Hanwell, M. D.; Curtis, D. E.; Lonie, D. C.; Vandermeersch, T.; Zurek, E.; Hutchison, G. R. Avogadro: an advanced semantic chemical editor, visualization, and analysis platform. J. Cheminf. 2012, 4, 17,  DOI: 10.1186/1758-2946-4-17
    Google Scholar
    67
    Avogadro: an advanced semantic chemical editor, visualization, and analysis platform
    Hanwell, Marcus D.; Curtis, Donald E.; Lonie, David C.; Vandermeersch, Tim; Zurek, Eva; Hutchison, Geoffrey R.
    Journal of Cheminformatics (2012), 4 (), 17CODEN: JCOHB3; ISSN:1758-2946. (Chemistry Central Ltd.)
    Background: The Avogadro project has developed an advanced mol. editor and visualizer designed for cross-platform use in computational chem., mol. modeling, bioinformatics, materials science, and related areas. It offers flexible, high quality rendering, and a powerful plugin architecture. Typical uses include building mol. structures, formatting input files, and analyzing output of a wide variety of computational chem. packages. By using the CML file format as its native document type, Avogadro seeks to enhance the semantic accessibility of chem. data types. Results: The work presented here details the Avogadro library, which is a framework providing a code library and application programming interface (API) with three-dimensional visualization capabilities; and has direct applications to research and education in the fields of chem., physics, materials science, and biol. The Avogadro application provides a rich graphical interface using dynamically loaded plugins through the library itself. The application and library can each be extended by implementing a plugin module in C++ or Python to explore different visualization techniques, build/manipulate mol. structures, and interact with other programs. We describe some example extensions, one which uses a genetic algorithm to find stable crystal structures, and one which interfaces with the PackMol program to create packed, solvated structures for mol. dynamics simulations. The 1.0 release series of Avogadro is the main focus of the results discussed here. Conclusions: Avogadro offers a semantic chem. builder and platform for visualization and anal. For users, it offers an easy-to-use builder, integrated support for downloading from common databases such as PubChem and the Protein Data Bank, extg. chem. data from a wide variety of formats, including computational chem. output, and native, semantic support for the CML file format. For developers, it can be easily extended via a powerful plugin mechanism to support new features in org. chem., inorg. complexes, drug design, materials, biomols., and simulations.
  68. 68
    Avogadro2 Version 1.97. https://two.avogadro.cc/.
    Google Scholar
    There is no corresponding record for this reference.
  69. 69
    Virtanen, P.; Gommers, R.; Oliphant, T. E.; Haberland, M.; Reddy, T.; Cournapeau, D.; Burovski, E.; Peterson, P.; Weckesser, W.; Bright, J. SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python. Nat. Methods 2020, 17, 261– 272,  DOI: 10.1038/s41592-019-0686-2
    Google Scholar
    69
    SciPy 1.0: fundamental algorithms for scientific computing in Python
    Virtanen, Pauli; Gommers, Ralf; Oliphant, Travis E.; Haberland, Matt; Reddy, Tyler; Cournapeau, David; Burovski, Evgeni; Peterson, Pearu; Weckesser, Warren; Bright, Jonathan; van der Walt, Stefan J.; Brett, Matthew; Wilson, Joshua; Millman, K. Jarrod; Mayorov, Nikolay; Nelson, Andrew R. J.; Jones, Eric; Kern, Robert; Larson, Eric; Carey, C. J.; Polat, Ilhan; Feng, Yu; Moore, Eric W.; Vander Plas, Jake; Laxalde, Denis; Perktold, Josef; Cimrman, Robert; Henriksen, Ian; Quintero, E. A.; Harris, Charles R.; Archibald, Anne M.; Ribeiro, Antonio H.; Pedregosa, Fabian; van Mulbregt, Paul
    Nature Methods (2020), 17 (3), 261-272CODEN: NMAEA3; ISSN:1548-7091. (Nature Research)
    Abstr.: SciPy is an open-source scientific computing library for the Python programming language. Since its initial release in 2001, SciPy has become a de facto std. for leveraging scientific algorithms in Python, with over 600 unique code contributors, thousands of dependent packages, over 100,000 dependent repositories and millions of downloads per yr. In this work, we provide an overview of the capabilities and development practices of SciPy 1.0 and highlight some recent tech. developments.
  70. 70
    Chan, L.; Hutchison, G. R.; Morris, G. M. Bayesian optimization for conformer generation. J. Cheminf. 2019, 11, 32,  DOI: 10.1186/s13321-019-0354-7
    Google Scholar
    There is no corresponding record for this reference.
  71. 71
    Chan, L.; Hutchison, G. R.; Morris, G. M. BOKEI: Bayesian optimization using knowledge of correlated torsions and expected improvement for conformer generation. Phys. Chem. Chem. Phys. 2020, 22, 5211– 5219,  DOI: 10.1039/C9CP06688H
    Google Scholar
    71
    BOKEI: Bayesian optimization using knowledge of correlated torsions and expected improvement for conformer generation
    Chan, Lucian; Hutchison, Geoffrey R.; Morris, Garrett M.
    Physical Chemistry Chemical Physics (2020), 22 (9), 5211-5219CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
    A key challenge in conformer sampling is finding low-energy conformations with a small no. of energy evaluations. We recently demonstrated the Bayesian Optimization Algorithm (BOA) is an effective method for finding the lowest energy conformation of a small mol. Our approach balances between exploitation and exploration, and is more efficient than exhaustive or random search methods. Here, we extend strategies used on proteins and oligopeptides (e.g. Ramachandran plots of secondary structure) and study correlated torsions in small mols. We use bivariate von Mises distributions to capture correlations, and use them to constrain the search space. We validate the performance of our new method, Bayesian Optimization with Knowledge-based Expected Improvement (BOKEI), on a dataset consisting of 533 diverse small mols., using (i) a force field (MMFF94); and (ii) a semi-empirical method (GFN2), as the objective function. We compare the search performance of BOKEI, BOA with Expected Improvement (BOA-EI), and a genetic algorithm (GA), using a fixed no. of energy evaluations. In more than 60% of the cases examd., BOKEI finds lower energy conformations than global optimization with BOA-EI or GA. More importantly, we find correlated torsions in up to 15% of small mols. in larger data sets, up to 8 times more often than previously reported. The BOKEI patterns not only describe steric clashes, but also reflect favorable intramol. interactions such as hydrogen bonds and π-π stacking. Increasing our understanding of the conformational preferences of mols. will help improve our ability to find low energy conformers efficiently, which will have impact in a wide range of computational modeling applications.
  72. 72
    Pordes, R.; Petravick, D.; Kramer, B.; Olson, D.; Livny, M.; Roy, A.; Avery, P.; Blackburn, K.; Wenaus, T.; Würthwein, F. The Open Science Grid. J. Phys. Conf. 2007, 78, 012057,  DOI: 10.1088/1742-6596/78/1/012057
    Google Scholar
    There is no corresponding record for this reference.
  73. 73
    Sfiligoi, I.; Bradley, D. C.; Holzman, B.; Mhashilkar, P.; Padhi, S.; Wurthwein, F. The Pilot Way to Grid Resources Using glideinWMS. World Congr. Comput. Sci. Inf. Eng. 2009, 2, 428– 432,  DOI: 10.1109/CSIE.2009.950
    Google Scholar
    There is no corresponding record for this reference.

Cited By

Click to copy section linkSection link copied!
Citation Statements
Explore this article's citation statements on scite.ai
powered by  Scite

This article is cited by 3 publications.

  1. The-Chuong Trinh, Pierre Falson, Viet-Khoa Tran-Nguyen, Ahcène Boumendjel. Ligand-Based Drug Discovery Leveraging State-of-the-Art Machine Learning Methodologies Exemplified by Cdr1 Inhibitor Prediction. Journal of Chemical Information and Modeling 2025, 65 (8) , 4027-4042. https://doi.org/10.1021/acs.jcim.5c00374
  2. Linghan Kong, Richard A. Bryce. Discriminating High from Low Energy Conformers of Druglike Molecules: An Assessment of Machine Learning Potentials and Quantum Chemical Methods. ChemPhysChem 2025, 26 (8) https://doi.org/10.1002/cphc.202400992
  3. Philipp Pracht, Stefan Grimme, Christoph Bannwarth, Fabian Bohle, Sebastian Ehlert, Gereon Feldmann, Johannes Gorges, Marcel Müller, Tim Neudecker, Christoph Plett, Sebastian Spicher, Pit Steinbach, Patryk A. Wesołowski, Felix Zeller. CREST—A program for the exploration of low-energy molecular chemical space. The Journal of Chemical Physics 2024, 160 (11) https://doi.org/10.1063/5.0197592
Download PDF
Go to Journal of Chemical Information and Modeling
Get e-Alerts
Get e-Alerts

Journal of Chemical Information and Modeling

Cite this: J. Chem. Inf. Model. 2023, 63, 23, 7401–7411
Click to copy citationCitation copied!
https://doi.org/10.1021/acs.jcim.3c01278
Published November 24, 2023

Copyright © 2023 The Authors. Published by American Chemical Society. This publication is licensed under

CC-BY 4.0 .

Article Views

1500

Altmetric

-

Citations

3
Learn about these metrics

Article Views are the COUNTER-compliant sum of full text article downloads since November 2008 (both PDF and HTML) across all institutions and individuals. These metrics are regularly updated to reflect usage leading up to the last few days.

Citations are the number of other articles citing this article, calculated by Crossref and updated daily. Find more information about Crossref citation counts.

The Altmetric Attention Score is a quantitative measure of the attention that a research article has received online. Clicking on the donut icon will load a page at altmetric.com with additional details about the score and the social media presence for the given article. Find more information on the Altmetric Attention Score and how the score is calculated.

  • Abstract

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 1

    Figure 1. Comparison of the smallest non-hydrogen RMSD between experimental crystallographic geometry and CREST or ETKDG conformers for the (left) Crystallographic Open Database (COD) and (right) Platinum Diverse data set. (c–f) Captions indicate best-fit linear regression in Å. Note that for both data sets, CREST produces smaller RMSD for molecules with few rotatable bonds but a larger slope indicating generally worse RMSD for larger compounds with more rotatable bonds.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 2

    Figure 2. Calculated radius of gyration for the lowest-energy CREST conformer and experimental geometries as a function of the number of rotatable bonds for the (a) Crystallographic Open Database (COD) and (b) Platinum data sets and scatterplot of compounds from (c) COD and (d) Platinum sets, comparing the radius of gyration from the CREST/GFN2 lowest-energy conformation with that of the experimental crystal structure geometry. Dashed line indicates a 1:1 correspondence, with approximate bounds indicated by solid lines.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 3

    Figure 3. Correlation between experimental and gas-phase torsions across the COD data set for (a) acyclic patterns and (b) ring patterns.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 4

    Figure 4. Histograms for pattern 229, including COD experimental torsional angles, gas-phase lowest-energy conformers from the same COD molecules, and gas-phase lowest-energy conformers across the entire data set, indicating the strong correlations and that the increased quantity of data greatly refines the histograms.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 5

    Figure 5. Histograms for pattern 307, including COD experimental torsional angles, gas-phase lowest-energy conformers from the same COD molecules, and gas-phase lowest-energy conformers across the entire data set, again indicating how the increased quantity of data greatly refines the torsional preferences (e.g., peaks near 50, 90, and 132°).

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 6

    Figure 6. Histograms for COD experimental torsional angles, gas-phase lowest-energy conformers from the same COD molecules, and conformers across the entire data set, again indicating that the increased quantity of data greatly refines the torsional preferences (e.g., a strong peak at 60°).

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 7

    Figure 7. Differences in torsional preferences for experimental and gas-phase geometries.

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 8

    Figure 8. Example of compound matching torsion pattern 270 with steric constraint forcing an angle of ∼90°. Figure from Avogadro2. (67,68)

    High Resolution Image
    Download MS PowerPoint Slide

    Figure 9

    Figure 9. Histograms of correlation between (a,d) ETKDG, (b,e) cosine fits, or (c,f) Gaussian fits and derived torsional histograms for (a–c) acyclic and (d–f) ring torsional patterns.

    High Resolution Image
    Download MS PowerPoint Slide

  • References

    This article references 73 other publications.

    1. 1
      Hawkins, P. C. D. Conformation Generation: The State of the Art. J. Chem. Inf. Model. 2017, 57, 1747– 1756,  DOI: 10.1021/acs.jcim.7b00221
      1
      Conformation Generation: The State of the Art
      Hawkins, Paul C. D.
      Journal of Chemical Information and Modeling (2017), 57 (8), 1747-1756CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The generation of conformations for small mols. is a problem of continuing interest in cheminformatics and computational drug discovery. This review will present an overview of methods used to sample conformational space, focusing on those methods designed for org. mols. commonly of interest in drug discovery. Different approaches to both the sampling of conformational space and the scoring of conformational stability will be compared and contrasted, with an emphasis on those methods suitable for conformer sampling of large nos. of drug-like mols. Particular attention will be devoted to the appropriate utilization of information from exptl. solid-state structures in validating and evaluating the performance of these tools. The review will conclude with some areas worthy of further investigation.
    2. 2
      Friedrich, N.-O.; de Bruyn Kops, C.; Flachsenberg, F.; Sommer, K.; Rarey, M.; Kirchmair, J. Benchmarking Commercial Conformer Ensemble Generators. J. Chem. Inf. Model. 2017, 57, 2719– 2728,  DOI: 10.1021/acs.jcim.7b00505
      2
      Benchmarking Commercial Conformer Ensemble Generators
      Friedrich, Nils-Ole; de Bruyn Kops, Christina; Flachsenberg, Florian; Sommer, Kai; Rarey, Matthias; Kirchmair, Johannes
      Journal of Chemical Information and Modeling (2017), 57 (11), 2719-2728CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      We assess and compare the performance of eight com. conformer ensemble generators (ConfGen, ConfGenX, cxcalc, iCon, MOE LowModeMD, MOE Stochastic, MOE Conformation Import, and OMEGA) and one leading free algorithm, the distance geometry algorithm implemented in RDKit. The comparative study is based on a new version of the Platinum Diverse Dataset, a high-quality benchmarking dataset of 2859 protein-bound ligand conformations extd. from the PDB. Differences in the performance of com. algorithms are much smaller than those obsd. for free algorithms in our previous study. For com. algorithms, the median min. root-mean-square deviations measured between protein-bound ligand conformations and ensembles of a max. of 250 conformers are between 0.46 and 0.61 Å. Com. conformer ensemble generators are characterized by their high robustness, with at least 99% of all input mols. successfully processed and few or even no substantial geometrical errors detectable in their output conformations. The RDKit distance geometry algorithm (with minimization enabled) appears to be a good free alternative since its performance is comparable to that of the midranked com. algorithms. Based on a statistical anal., we elaborate on which algorithms to use and how to parametrize them for best performance in different application scenarios.
    3. 3
      Friedrich, N.-O.; Meyder, A.; de Bruyn Kops, C.; Sommer, K.; Flachsenberg, F.; Rarey, M.; Kirchmair, J. High-Quality Dataset of Protein-Bound Ligand Conformations and Its Application to Benchmarking Conformer Ensemble Generators. J. Chem. Inf. Model. 2017, 57, 529– 539,  DOI: 10.1021/acs.jcim.6b00613
      3
      High-Quality Dataset of Protein-Bound Ligand Conformations and Its Application to Benchmarking Conformer Ensemble Generators
      Friedrich, Nils-Ole; Meyder, Agnes; de Bruyn Kops, Christina; Sommer, Kai; Flachsenberg, Florian; Rarey, Matthias; Kirchmair, Johannes
      Journal of Chemical Information and Modeling (2017), 57 (3), 529-539CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The authors developed a cheminformatics pipeline for the fully automated selection and extn. of high-quality protein-bound ligand conformations from x-ray structural data. The pipeline evaluates the validity and accuracy of the 3D structures of small mols. according to multiple criteria, including their fit to the electron d. and their physicochem. and structural properties. Using this approach, the authors compiled two high-quality datasets from the Protein Data Bank (PDB): a comprehensive dataset and a diversified subset of 4626 and 2912 structures, resp. The datasets were applied to benchmarking seven freely available conformer ensemble generators: Balloon (two different algorithms), the RDKit std. conformer ensemble generator, the Exptl.-Torsion basic Knowledge Distance Geometry (ETKDG) algorithm, Confab, Frog2 and Multiconf-DOCK. Substantial differences in the performance of the individual algorithms were obsd., with RDKit and ETKDG generally achieving a favorable balance of accuracy, ensemble size and runtime. The platinum datasets are available for download from http://www.zbh.uni-hamburg.de/platinum_dataset.
    4. 4
      Ebejer, J.-P.; Morris, G. M.; Deane, C. M. Freely Available Conformer Generation Methods: How Good Are They?. J. Chem. Inf. Model. 2012, 52, 1146– 1158,  DOI: 10.1021/ci2004658
      4
      Freely Available Conformer Generation Methods: How Good Are They?
      Ebejer, Jean-Paul; Morris, Garrett M.; Deane, Charlotte M.
      Journal of Chemical Information and Modeling (2012), 52 (5), 1146-1158CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      A review. Conformer generation has important implications in cheminformatics, particularly in computational drug discovery where the quality of conformer generation software may affect the outcome of a virtual screening exercise. We examine the performance of four freely available small mol. conformer generation tools (Balloon, Confab, Frog2, and RDKit) alongside a com. tool (MOE). The aim of this study is 3-fold: (i) to identify which tools most accurately reproduce exptl. detd. structures; (ii) to examine the diversity of the generated conformational set; and (iii) to benchmark the computational time expended. These aspects were tested using a set of 708 drug-like mols. assembled from the OMEGA validation set and the Astex Diverse Set. These mols. have varying physicochem. properties and at least one known X-ray crystal structure. We found that RDKit and Confab are statistically better than other methods at generating low rmsd conformers to the known structure. RDKit is particularly suited for less flexible mols. while Confab, with its systematic approach, is able to generate conformers which are geometrically closer to the exptl. detd. structure for mols. with a large no. of rotatable bonds (≥10). In our tests RDKit also resulted as the second fastest method after Frog2. In order to enhance the performance of RDKit, we developed a postprocessing algorithm to build a diverse and representative set of conformers which also contains a close conformer to the known structure. Our anal. indicates that, with postprocessing, RDKit is a valid free alternative to com., proprietary software.
    5. 5
      Penner, P.; Guba, W.; Schmidt, R.; Meyder, A.; Stahl, M.; Rarey, M. The Torsion Library: Semiautomated Improvement of Torsion Rules with SMARTScompare. J. Chem. Inf. Model. 2022, 62, 1644– 1653,  DOI: 10.1021/acs.jcim.2c00043
      5
      The Torsion Library: Semiautomated Improvement of Torsion Rules with SMARTScompare
      Penner, Patrick; Guba, Wolfgang; Schmidt, Robert; Meyder, Agnes; Stahl, Martin; Rarey, Matthias
      Journal of Chemical Information and Modeling (2022), 62 (7), 1644-1653CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The Torsion Library is a collection of torsion motifs assocd. with angle distributions, derived from crystallog. databases. It is used in strain assessment, conformer generation, and geometry optimization. A hierarchical structure of expert curated SMARTS defines the chem. environments of rotatable bonds and assocs. these with preferred angles. SMARTS can be very complex and full of implications, which make them difficult to maintain manually. Recent developments in automatically comparing SMARTS patterns can be applied to the Torsion Library to ensure its correctness. We specifically discuss the implementation and the limits of such a procedure in the context of torsion motifs and show several examples of how the Torsion Library benefits from this. All automated changes are validated manually and then shown to have an effect on the angle distributions by correcting matching behavior. The cor. Torsion Library itself is available including both PDB as well as CSD histograms in the Supporting Information and can be used to evaluate rotatable bonds at https://torsions.zbh.uni-hamburg.de.
    6. 6
      Wang, S.; Witek, J.; Landrum, G. A.; Riniker, S. Improving Conformer Generation for Small Rings and Macrocycles Based on Distance Geometry and Experimental Torsional-Angle Preferences. J. Chem. Inf. Model. 2020, 60, 2044– 2058,  DOI: 10.1021/acs.jcim.0c00025
      6
      Improving Conformer Generation for Small Rings and Macrocycles Based on Distance Geometry and Experimental Torsional-Angle Preferences
      Wang, Shuzhe; Witek, Jagna; Landrum, Gregory A.; Riniker, Sereina
      Journal of Chemical Information and Modeling (2020), 60 (4), 2044-2058CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The conformer generator ETKDG is a stochastic search method that utilizes distance geometry together with knowledge derived from exptl. crystal structures. It has been shown to generate good conformers for acyclic, flexible mols. This work builds on ETKDG to improve conformer generation of mols. contg. small or large aliph. (i.e., non-arom.) rings. For one, we devise addnl. torsional-angle potentials to describe small aliph. rings and adapt the previously developed potentials for acyclic bonds to facilitate the sampling of macrocycles. However, due to the larger no. of degrees of freedom of macrocycles, the conformational space to sample is much broader than for small mols., creating a challenge for conformer generators. We therefore introduce different heuristics to restrict the search space of macrocycles and bias the sampling toward more exptl. relevant structures. Specifically, we show the usage of elliptical geometry and customizable Coulombic interactions as heuristics. The performance of the improved ETKDG is demonstrated on test sets of diverse macrocycles and cyclic peptides. The code developed here will be incorporated into the 2020.03 release of the open-source cheminformatics library RDKit.
    7. 7
      Riniker, S.; Landrum, G. A. Better Informed Distance Geometry: Using What We Know To Improve Conformation Generation. J. Chem. Inf. Model. 2015, 55, 2562– 2574,  DOI: 10.1021/acs.jcim.5b00654
      7
      Better Informed Distance Geometry: Using What We Know To Improve Conformation Generation
      Riniker, Sereina; Landrum, Gregory A.
      Journal of Chemical Information and Modeling (2015), 55 (12), 2562-2574CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      Small org. mols. are often flexible, i.e., they can adopt a variety of low-energy conformations in soln. that exist in equil. with each other. Two main search strategies are used to generate representative conformational ensembles for mols.: systematic and stochastic. In the first approach, each rotatable bond is sampled systematically in discrete intervals, limiting its use to mols. with a small no. of rotatable bonds. Stochastic methods, however, sample the conformational space of a mol. randomly and can thus be applied to more flexible mols. Different methods employ different degrees of exptl. data for conformer generation. So-called knowledge-based methods use predefined libraries of torsional angles and ring conformations. In the distance geometry approach, however, a smaller amt. of empirical information was used, i.e., ideal bond lengths, ideal bond angles, and a few ideal torsional angles. Distance geometry is a computationally fast method to generate conformers, but it has the downside that purely distance-based constraints tend to lead to distorted arom. rings and sp2 centers. To correct this, the resulting conformations are often minimized with a force field, adding computational complexity and run time. Here the authors present an alternative strategy that combines the distance geometry approach with exptl. torsion-angle preferences obtained from small-mol. crystallog. data. The torsional angles are described by a previously developed set of hierarchically structured SMARTS patterns. The new approach is implemented in the open-source cheminformatics library RDKit, and its performance is assessed by comparing the diversity of the generated ensemble and the ability to reproduce crystal conformations taken from the crystal structures of small mols. and protein-ligand complexes.
    8. 8
      Guba, W.; Meyder, A.; Rarey, M.; Hert, J. Torsion Library Reloaded: A New Version of Expert-Derived SMARTS Rules for Assessing Conformations of Small Molecules. J. Chem. Inf. Model. 2016, 56, 1– 5,  DOI: 10.1021/acs.jcim.5b00522
      8
      Torsion Library Reloaded: A New Version of Expert-Derived SMARTS Rules for Assessing Conformations of Small Molecules
      Guba, Wolfgang; Meyder, Agnes; Rarey, Matthias; Hert, Jerome
      Journal of Chemical Information and Modeling (2016), 56 (1), 1-5CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The Torsion Library contains hundreds of rules for small mol. conformations which have been derived from the Cambridge Structural Database (CSD) and are curated by mol. design experts. The torsion rules are encoded as SMARTS patterns and categorize rotatable bonds via a traffic light coloring scheme. We have systematically revised all torsion rules to better identify highly strained conformations and minimize the no. of false alerts for CSD small mol. X-ray structures. For this new release, we added or substantially modified 78 torsion patterns and reviewed all angles and tolerance intervals. The overall no. of red alerts for a filtered CSD data set with 130 000 structures was reduced by a factor of 4 compared to the predecessor. This is of clear advantage in 3D virtual screening where hits should only be removed by a conformational filter if they are in energetically inaccessible conformations.
    9. 9
      Gražulis, S.; Chateigner, D.; Downs, R. T.; Yokochi, A. F. T.; Quirós, M.; Lutterotti, L.; Manakova, E.; Butkus, J.; Moeck, P.; Le Bail, A. Crystallography Open Database – an open-access collection of crystal structures. J. Appl. Crystallogr. 2009, 42, 726– 729,  DOI: 10.1107/S0021889809016690
      9
      Crystallography Open Database - an open-access collection of crystal structures
      Grazulis, Saulius; Chateigner, Daniel; Downs, Robert T.; Yokochi, A. F. T.; Quiros, Miguel; Lutterotti, Luca; Manakova, Elena; Butkus, Justas; Moeck, Peter; Le Bail, Armel
      Journal of Applied Crystallography (2009), 42 (4), 726-729CODEN: JACGAR; ISSN:0021-8898. (International Union of Crystallography)
      The Crystallog. Open Database (COD), which is a project that aims to gather all available inorg., metal-org. and small org. mol. structural data in one database, is described. The database adopts an open-access model. The COD currently contains ∼80,000 entries in crystallog. information file format, with nearly full coverage of the International Union of Crystallog. publications, and is growing in size and quality.
    10. 10
      Gražulis, S.; Daškevič, A.; Merkys, A.; Chateigner, D.; Lutterotti, L.; Quirós, M.; Serebryanaya, N. R.; Moeck, P.; Downs, R. T.; Le Bail, A. Crystallography Open Database (COD): an open-access collection of crystal structures and platform for world-wide collaboration. Nucleic Acids Res. 2012, 40, D420– D427,  DOI: 10.1093/nar/gkr900
      10
      Crystallography Open Database (COD): an open-access collection of crystal structures and platform for world-wide collaboration
      Grazulis, Saulius; Daskevic, Adriana; Merkys, Andrius; Chateigner, Daniel; Lutterotti, Luca; Quiros, Miguel; Serebryanaya, Nadezhda R.; Moeck, Peter; Downs, Robert T.; Le Bail, Armel
      Nucleic Acids Research (2012), 40 (D1), D420-D427CODEN: NARHAD; ISSN:0305-1048. (Oxford University Press)
      Using an open-access distribution model, the Crystallog. Open Database (COD, http://www.crystallog.net) collects all known small mol. / small to medium sized unit cell' crystal structures and makes them available freely on the Internet. As of today, the COD has aggregated ∼150 000 structures, offering basic search capabilities and the possibility to download the whole database, or parts thereof using a variety of std. open communication protocols. A newly developed website provides capabilities for all registered users to deposit published and so far unpublished structures as personal communications or pre-publication depositions. Such a setup enables extension of the COD database by many users simultaneously. This increases the possibilities for growth of the COD database, and is the first step towards establishing a world wide Internet-based collaborative platform dedicated to the collection and curation of structural knowledge.
    11. 11
      Groom, C. R.; Bruno, I. J.; Lightfoot, M. P.; Ward, S. C. The Cambridge Structural Database. Acta Crystallogr. B 2016, 72, 171– 179,  DOI: 10.1107/S2052520616003954
      11
      The Cambridge Structural Database
      Groom, Colin R.; Bruno, Ian J.; Lightfoot, Matthew P.; Ward, Suzanna C.
      Acta Crystallographica, Section B: Structural Science, Crystal Engineering and Materials (2016), 72 (2), 171-179CODEN: ACSBDA; ISSN:2052-5206. (International Union of Crystallography)
      The Cambridge Structural Database (CSD) contains a complete record of all published org. and metal-org. small-mol. crystal structures. The database has been in operation for over 50 years and continues to be the primary means of sharing structural chem. data and knowledge across disciplines. As well as structures that are made public to support scientific articles, it includes many structures published directly as CSD Communications. All structures are processed both computationally and by expert structural chem. editors prior to entering the database. A key component of this processing is the reliable assocn. of the chem. identity of the structure studied with the exptl. data. This important step helps ensure that data is widely discoverable and readily reusable. Content is further enriched through selective inclusion of addnl. exptl. data. Entries are available to anyone through free CSD community web services. Linking services developed and maintained by the CCDC, combined with the use of std. identifiers, facilitate discovery from other resources. Data can also be accessed through CCDC and third party software applications and through an application programming interface.
    12. 12
      Sadowski, P.; Baldi, P. Small-Molecule 3D Structure Prediction Using Open Crystallography Data. J. Chem. Inf. Model. 2013, 53, 3127– 3130,  DOI: 10.1021/ci4005282
      12
      Small-Molecule 3D Structure Prediction Using Open Crystallography Data
      Sadowski, Peter; Baldi, Pierre
      Journal of Chemical Information and Modeling (2013), 53 (12), 3127-3130CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      Predicting the 3D structures of small mols. is a common problem in chemoinformatics. Even the best methods are inaccurate for complex mols., and there is a large gap in accuracy between proprietary and free algorithms. Previous work presented COSMOS, a novel data-driven algorithm that uses knowledge of known structures from the Cambridge Structural Database and demonstrates performance that was competitive with proprietary algorithms. However, dependence on the Cambridge Structural Database prevented its widespread use. Here, we present an updated version of the COSMOS structure predictor, complete with a free structure library derived from open data sources. We demonstrate that COSMOS performs better than other freely available methods, with a mean RMSD of 1.16 and 1.68 Å for org. and metal-org. structures, resp., and a mean prediction time of 60 ms per mol. This is a 17% and 20% redn., resp., in RMSD compared to the free predictor provided by Open Babel, and it is 10 times faster. The ChemDB Web portal provides a COSMOS prediction Web server, as well as downloadable copies of the COSMOS executable and library of mol. substructures.
    13. 13
      Wicker, J. G. P.; Cooper, R. I. Beyond Rotatable Bond Counts: Capturing 3D Conformational Flexibility in a Single Descriptor. J. Chem. Inf. Model. 2016, 56, 2347– 2352,  DOI: 10.1021/acs.jcim.6b00565
      13
      Beyond Rotatable Bond Counts: Capturing 3D Conformational Flexibility in a Single Descriptor
      Wicker, Jerome G. P.; Cooper, Richard I.
      Journal of Chemical Information and Modeling (2016), 56 (12), 2347-2352CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      A new mol. descriptor, nConf20, based on chem. connectivity, is presented which captures the accessible conformational space of a mol. Currently the best available two-dimensional descriptors for quantifying the flexibility of a particular mol. are the rotatable bond count (RBC) and the Kier flexibility index. The authors present a descriptor which captures this information by sampling the conformational space of a mol. using the RDKit conformer generator. Flexibility has previously been identified as a key feature in detg. whether a mol. is likely to crystallize or not. For this application, nConf20 significantly outperforms previously reported single-variable classifiers and also assists rule-based anal. of black-box machine learning classification algorithms.
    14. 14
      Das, S.; Dinpazhoh, L.; Tanemura, K. A.; Merz, K. M. Rapid and Automated Ab Initio Metabolite Collisional Cross Section Prediction from SMILES Input. J. Chem. Inf. Model. 2023, 63, 4995– 5000,  DOI: 10.1021/acs.jcim.3c00890
      14
      Rapid and Automated Ab Initio Metabolite Collisional Cross Section Prediction from SMILES Input
      Das, Susanta; Dinpazhoh, Laleh; Tanemura, Kiyoto Aramis; Merz Jr., Kenneth M.
      Journal of Chemical Information and Modeling (2023), 63 (16), 4995-5000CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      We implemented an ab initio CCS prediction workflow which incrementally refines generated structures using mol. mechanics, a deep learning potential, conformational clustering, and quantum mechanics (QM). Automating intermediate steps for a high performance computing (HPC) environment allows users to input the SMILES structure of small org. mols. and obtain a Boltzmann averaged collisional cross section (CCS) value as output. The CCS of a mol. species is a metric measured by ion mobility spectrometry (IMS) which can improve annotation of untargeted metabolomics expts. We report only a minor drop in accuracy when we expedite the CCS calcn. by replacing the QM geometry refinement step with a single-point energy calcn. Even though the workflow involves stochastic steps (i.e., conformation generation and clustering), the final CCS value was highly reproducible for multiple iterations on L-carnosine. Finally, we illustrate that the gas phase ensemble modeled for the workflow are intermediate files which can be used for the prediction of other properties such as aq. phase NMR chem. shift prediction.
    15. 15
      Das, S.; Tanemura, K. A.; Dinpazhoh, L.; Keng, M.; Schumm, C.; Leahy, L.; Asef, C. K.; Rainey, M.; Edison, A. S.; Fernández, F. M.; Merz, K. M. In Silico Collision Cross Section Calculations to Aid Metabolite Annotation. J. Am. Soc. Mass Spectrom. 2022, 33, 750– 759,  DOI: 10.1021/jasms.1c00315
      15
      In Silico Collision Cross Section Calculations to Aid Metabolite Annotation
      Das, Susanta; Tanemura, Kiyoto Aramis; Dinpazhoh, Laleh; Keng, Mithony; Schumm, Christina; Leahy, Lydia; Asef, Carter K.; Rainey, Markace; Edison, Arthur S.; Fernandez, Facundo M.; Merz Jr., Kenneth M.
      Journal of the American Society for Mass Spectrometry (2022), 33 (5), 750-759CODEN: JAMSEF; ISSN:1879-1123. (American Chemical Society)
      The interpretation of ion mobility coupled to mass spectrometry (IM-MS) data to predict unknown structures is challenging and depends on accurate theor. ests. of the mol. ion collision cross section (CCS) against a buffer gas in a low or atm. pressure drift chamber. The sensitivity and reliability of computational prediction of CCS values depend on accurately modeling the mol. state over accessible conformations. In this work, we developed an efficient CCS computational workflow using a machine learning model in conjunction with std. DFT methods and CCS calcns. Furthermore, we have performed Traveling Wave IM-MS (TWIMS) expts. to validate the extant exptl. values and assess uncertainties in exptl. measured CCS values. The developed workflow yielded accurate structural predictions and provides unique insights into the likely preferred conformation analyzed using IM-MS expts. The complete workflow makes the computation of CCS values tractable for a large no. of conformationally flexible metabolites with complex mol. structures.
    16. 16
      Insausti, A.; Alonso, E. R.; Tercero, B.; Santos, J. I.; Calabrese, C.; Vogt, N.; Corzana, F.; Demaison, J.; Cernicharo, J.; Cocinero, E. J. Laboratory Observation of, Astrochemical Search for, and Structure of Elusive Erythrulose in the Interstellar Medium. J. Phys. Chem. Lett. 2021, 12, 1352– 1359,  DOI: 10.1021/acs.jpclett.0c03050
      16
      Laboratory Observation of, Astrochemical Search for, and Structure of Elusive Erythrulose in the Interstellar Medium
      Insausti, Aran; Alonso, Elena R.; Tercero, Belen; Santos, Jose I.; Calabrese, Camilla; Vogt, Natalja; Corzana, Francisco; Demaison, Jean; Cernicharo, Jose; Cocinero, Emilio J.
      Journal of Physical Chemistry Letters (2021), 12 (4), 1352-1359CODEN: JPCLCD; ISSN:1948-7185. (American Chemical Society)
      Rotational spectroscopy provides the most powerful means of identifying mols. of biol. interest in the interstellar medium (ISM), but despite their importance, the detection of carbohydrates has remained rather elusive. Here, we present a comprehensive Fourier transform rotational spectroscopic study of elusive erythrulose, a sugar building block likely to be present in the ISM, employing a novel method of transferring the hygroscopic oily carbohydrate into the gas phase. The high sensitivity of the expt. allowed the rotational spectra of all monosubstituted isotopologue species of 13C-12C3H8O4 to be recorded, which, together with quantum chem. calcns., enabled us to det. their equil. geometries (reSE) with great precision. Searches employing the new exptl. data for erythrulose have been undertaken in different ISM regions, so far including the cold areas Barnard 1, the pre-stellar core TMC-1, Sagittarius B2. Although no lines of erythrulose were found, this data will serve to enable future searches and possible detections in other ISM regions.
    17. 17
      Alonso, E. R.; Peña, I.; Cabezas, C.; Alonso, J. L. Structural Expression of Exo-Anomeric Effect. J. Phys. Chem. Lett. 2016, 7, 845– 850,  DOI: 10.1021/acs.jpclett.6b00028
      17
      Structural Expression of Exo-Anomeric Effect
      Alonso, Elena R.; Pena, Isabel; Cabezas, Carlos; Alonso, Jose L.
      Journal of Physical Chemistry Letters (2016), 7 (5), 845-850CODEN: JPCLCD; ISSN:1948-7185. (American Chemical Society)
      Structural signatures for exo-anomeric effect have been extd. from the archetypal Me β-D-xyloside using broadband Fourier transform microwave spectroscopy combined with laser ablation. Spectrum anal. allows the detn. of a set of rotational consts., which has been unequivocally attributed to conformer cc-β-4C1 g-, corresponding to the global min. of the potential energy surface, where the aglycon residue (CH3) orientation contributes to maximization of the exo-anomeric effect. Further anal. allowed the detn. of the rs structure, based on the detection of 11 isotopologues-derived from the presence of six 13C and five 18O atoms-obsd. in their natural abundances. The obsd. glycosidic C1-O1 bond length decrease (1.38 Å) can be interpreted in terms of the exo-anomeric effect. As such, the exo-anomeric effect presents itself as one of the main driving forces controlling the shape of many biol. important oligosaccharides.
    18. 18
      Peña, I.; Cocinero, E. J.; Cabezas, C.; Lesarri, A.; Mata, S.; Écija, P.; Daly, A. M.; Cimas, A.; Bermúdez, C.; Basterretxea, F. J.; Blanco, S.; Fernández, J. A.; López, J. C.; Castaño, F.; Alonso, J. L. Six Pyranoside Forms of Free 2-Deoxy-D-ribose. Angew. Chem., Int. Ed. 2013, 52, 11840– 11845,  DOI: 10.1002/anie.201305589
      18
      Six Pyranoside Forms of Free 2-Deoxy-D-ribose
      Pena, Isabel; Cocinero, Emilio J.; Cabezas, Carlos; Lesarri, Alberto; Mata, Santiago; Ecija, Patricia; Daly, Adam M.; Cimas, Alvaro; Bermudez, Celina; Basterretxea, Francisco J.; Blanco, Susana; Fernandez, Jose A.; Lopez, Juan C.; Castano, Fernando; Alonso, Jose L.
      Angewandte Chemie, International Edition (2013), 52 (45), 11840-11845CODEN: ACIEF5; ISSN:1433-7851. (Wiley-VCH Verlag GmbH & Co. KGaA)
      The gas phase rotational spectrum of 2-deoxy-D-ribose (I) was obsd. using a UV ultrafast laser ablation technique as a source for Balle-Flygare and chirped pulsed FTMW spectrometers and was assigned with the aide of MP2(FULL) calcns. In the gas phase, I exists as a mixt. of two α- (10%) and four β-pyranose (90%) conformers. the conformational behavior is controlled by both anomeric effects and hydrogen bonding. the OH groups are preferentially oriented so as to favor cooperative hydrogen bonding. In this context, the previous exptl. ionization potential of 9.1 eV of gas phase I (obtained using tunable vacuum UV synchrotron radiation ) assigned to the α-pyranose should actually correspond to the β-pyranose. the solvent effect on furanose-pyranose equil. and the occurrence of these conformers in RNA and DNA were discussed.
    19. 19
      Baldi, P. Data-Driven High-Throughput Prediction of the 3-D Structure of Small Molecules: Review and Progress. A Response to the Letter by the Cambridge Crystallographic Data Centre. J. Chem. Inf. Model. 2011, 51, 3029,  DOI: 10.1021/ci200460z
      19
      Data-Driven High-Throughput Prediction of the 3-D Structure of Small Molecules: Review and Progress. A Response to the Letter by the Cambridge Crystallographic Data Centre
      Baldi, Pierre
      Journal of Chemical Information and Modeling (2011), 51 (12), 3029CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      A polemic in response to Groom, C.R. (ibid., S1, 2787 - 2787, 2011) is presented to sentiments expressed in "Data-Driven High-Throughput Prediction of the 3-D Structure of Small Mols.: Review and Progress. A Response from The Cambridge Crystallog. Data Center", recently published in the Journal of Chem. Information and Modeling, which may give readers a misleading impression regarding significant impediments to scientific research posed by the CCDC.
    20. 20
      Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A.; Skiff, W. M. UFF, a full periodic table force field for molecular mechanics and molecular dynamics simulations. J. Am. Chem. Soc. 1992, 114, 10024– 10035,  DOI: 10.1021/ja00051a040
      20
      UFF, a full periodic table force field for molecular mechanics and molecular dynamics simulations
      Rappe, A. K.; Casewit, C. J.; Colwell, K. S.; Goddard, W. A., III; Skiff, W. M.
      Journal of the American Chemical Society (1992), 114 (25), 10024-35CODEN: JACSAT; ISSN:0002-7863.
      A new mol. mechanics force field, the Universal force field (UFF), is described wherein the force field parameters are estd. using general rules based only on the element, its hybridization and its connectivity. The force field functional forms, parameters, and generating formulas for the full periodic table are presented.
    21. 21
      Folmsbee, D.; Hutchison, G. Assessing conformer energies using electronic structure and machine learning methods. Int. J. Quantum Chem. 2021, 121, e26381  DOI: 10.1002/qua.26381
      21
      Assessing conformer energies using electronic structure and machine learning methods
      Folmsbee, Dakota; Hutchison, Geoffrey
      International Journal of Quantum Chemistry (2021), 121 (1), e26381CODEN: IJQCB2; ISSN:0020-7608. (John Wiley & Sons, Inc.)
      A review. We have performed a large-scale evaluation of current computational methods, including conventional small-mol. force fields; semiempirical, d. functional, ab initio electronic structure methods; and current machine learning (ML) techniques to evaluate relative single-point energies. Using up to 10 local min. geometries across ∼700 mols., each optimized by B3LYP-D3BJ with single-point DLPNO-CCSD(T) triple-zeta energies, we consider over 6500 single points to compare the correlation between different methods for both relative energies and ordered rankings of min. We find that the current ML methods have potential and recommend methods at each tier of the accuracy-time tradeoff, particularly the recent GFN2 semiempirical method, the B97-3c d. functional approxn., and RI-MP2 for accurate conformer energies. The ANI family of ML methods shows promise, particularly the ANI-1ccx variant trained in part on coupled-cluster energies. Multiple methods suggest continued improvements should be expected in both performance and accuracy.
    22. 22
      Kanal, I. Y.; Keith, J. A.; Hutchison, G. R. A sobering assessment of small-molecule force field methods for low energy conformer predictions. Int. J. Quantum Chem. 2018, 118, e25512  DOI: 10.1002/qua.25512
      There is no corresponding record for this reference.
    23. 23
      Bannwarth, C.; Ehlert, S.; Grimme, S. GFN2-xTB - An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions. J. Chem. Theory Comput. 2019, 15, 1652– 1671,  DOI: 10.1021/acs.jctc.8b01176
      23
      GFN2-xTB-An Accurate and Broadly Parametrized Self-Consistent Tight-Binding Quantum Chemical Method with Multipole Electrostatics and Density-Dependent Dispersion Contributions
      Bannwarth, Christoph; Ehlert, Sebastian; Grimme, Stefan
      Journal of Chemical Theory and Computation (2019), 15 (3), 1652-1671CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
      An extended semiempirical tight-binding model is presented, which is primarily designed for the fast calcn. of structures and noncovalent interactions energies for mol. systems with roughly 1000 atoms. The essential novelty in this so-called GFN2-xTB method is the inclusion of anisotropic second order d. fluctuation effects via short-range damped interactions of cumulative at. multipole moments. Without noticeable increase in the computational demands, this results in a less empirical and overall more phys. sound method, which does not require any classical halogen or hydrogen bonding corrections and which relies solely on global and element-specific parameters (available up to radon, Z = 86). Moreover, the at. partial charge dependent D4 London dispersion model is incorporated self-consistently, which can be naturally obtained in a tight-binding picture from second order d. fluctuations. Fully anal. and numerically precise gradients (nuclear forces) are implemented. The accuracy of the method is benchmarked for a wide variety of systems and compared with other semiempirical methods. Along with excellent performance for the "target" properties, we also find lower errors for "off-target" properties such as barrier heights and mol. dipole moments. High computational efficiency along with the improved physics compared to it precursor GFN-xTB makes this method well-suited to explore the conformational space of mol. systems. Significant improvements are furthermore obsd. for various benchmark sets, which are prototypical for biomol. systems in aq. soln.
    24. 24
      Smith, J. S.; Isayev, O.; Roitberg, A. E. ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost. Chem. Sci. 2017, 8, 3192– 3203,  DOI: 10.1039/C6SC05720A
      24
      ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost
      Smith, J. S.; Isayev, O.; Roitberg, A. E.
      Chemical Science (2017), 8 (4), 3192-3203CODEN: CSHCCN; ISSN:2041-6520. (Royal Society of Chemistry)
      A review. Deep learning is revolutionizing many areas of science and technol., esp. image, text, and speech recognition. In this paper, we demonstrate how a deep neural network (NN) trained on quantum mech. (QM) DFT calcns. can learn an accurate and transferable potential for org. mols. We introduce ANAKIN-ME (Accurate NeurAl networK engINe for Mol. Energies) or ANI for short. ANI is a new method designed with the intent of developing transferable neural network potentials that utilize a highly-modified version of the Behler and Parrinello symmetry functions to build single-atom at. environment vectors (AEV) as a mol. representation. AEVs provide the ability to train neural networks to data that spans both configurational and conformational space, a feat not previously accomplished on this scale. We utilized ANI to build a potential called ANI-1, which was trained on a subset of the GDB databases with up to 8 heavy atoms in order to predict total energies for org. mols. contg. four atom types: H, C, N, and O. To obtain an accelerated but phys. relevant sampling of mol. potential surfaces, we also proposed a Normal Mode Sampling (NMS) method for generating mol. conformations. Through a series of case studies, we show that ANI-1 is chem. accurate compared to ref. DFT calcns. on much larger mol. systems (up to 54 atoms) than those included in the training data set.
    25. 25
      Smith, J. S.; Nebgen, B.; Lubbers, N.; Isayev, O.; Roitberg, A. E. Less is more: Sampling chemical space with active learning. J. Chem. Phys. 2018, 148, 241733,  DOI: 10.1063/1.5023802
      25
      Less is more: Sampling chemical space with active learning
      Smith, Justin S.; Nebgen, Ben; Lubbers, Nicholas; Isayev, Olexandr; Roitberg, Adrian E.
      Journal of Chemical Physics (2018), 148 (24), 241733/1-241733/10CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
      The development of accurate and transferable machine learning (ML) potentials for predicting mol. energetics is a challenging task. The process of data generation to train such ML potentials is a task neither well understood nor researched in detail. In this work, we present a fully automated approach for the generation of datasets with the intent of training universal ML potentials. It is based on the concept of active learning (AL) via Query by Committee (QBC), which uses the disagreement between an ensemble of ML potentials to infer the reliability of the ensemble's prediction. QBC allows the presented AL algorithm to automatically sample regions of chem. space where the ML potential fails to accurately predict the potential energy. AL improves the overall fitness of ANAKIN-ME (ANI) deep learning potentials in rigorous test cases by mitigating human biases in deciding what new training data to use. AL also reduces the training set size to a fraction of the data required when using naive random sampling techniques. To provide validation of our AL approach, we develop the COmprehensive Machine-learning Potential (COMP6) benchmark (publicly available on GitHub) which contains a diverse set of org. mols. Active learning-based ANI potentials outperform the original random sampled ANI-1 potential with only 10% of the data, while the final active learning-based model vastly outperforms ANI-1 on the COMP6 benchmark after training to only 25% of the data. Finally, we show that our proposed AL technique develops a universal ANI potential (ANI-1x) that provides accurate energy and force predictions on the entire COMP6 benchmark. This universal ML potential achieves a level of accuracy on par with the best ML potentials for single mols. or materials, while remaining applicable to the general class of org. mols. composed of the elements CHNO. (c) 2018 American Institute of Physics.
    26. 26
      Smith, J. S.; Nebgen, B. T.; Zubatyuk, R.; Lubbers, N.; Devereux, C.; Barros, K.; Tretiak, S.; Isayev, O.; Roitberg, A. E. Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning. Nat. Commun. 2019, 10, 2903,  DOI: 10.1038/s41467-019-10827-4
      26
      Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning
      Smith Justin S; Devereux Christian; Roitberg Adrian E; Smith Justin S; Nebgen Benjamin T; Zubatyuk Roman; Lubbers Nicholas; Barros Kipton; Tretiak Sergei; Smith Justin S; Lubbers Nicholas; Nebgen Benjamin T; Tretiak Sergei; Zubatyuk Roman; Isayev Olexandr
      Nature communications (2019), 10 (1), 2903 ISSN:.
      Computational modeling of chemical and biological systems at atomic resolution is a crucial tool in the chemist's toolset. The use of computer simulations requires a balance between cost and accuracy: quantum-mechanical methods provide high accuracy but are computationally expensive and scale poorly to large systems, while classical force fields are cheap and scalable, but lack transferability to new systems. Machine learning can be used to achieve the best of both approaches. Here we train a general-purpose neural network potential (ANI-1ccx) that approaches CCSD(T)/CBS accuracy on benchmarks for reaction thermochemistry, isomerization, and drug-like molecular torsions. This is achieved by training a network to DFT data then using transfer learning techniques to retrain on a dataset of gold standard QM calculations (CCSD(T)/CBS) that optimally spans chemical space. The resulting potential is broadly applicable to materials science, biology, and chemistry, and billions of times faster than CCSD(T)/CBS calculations.
    27. 27
      Devereux, C.; Smith, J. S.; Huddleston, K. K.; Barros, K.; Zubatyuk, R.; Isayev, O.; Roitberg, A. E. Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens. J. Chem. Theory Comput. 2020, 16, 4192– 4202,  DOI: 10.1021/acs.jctc.0c00121
      27
      Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens
      Devereux, Christian; Smith, Justin S.; Huddleston, Kate K.; Barros, Kipton; Zubatyuk, Roman; Isayev, Olexandr; Roitberg, Adrian E.
      Journal of Chemical Theory and Computation (2020), 16 (7), 4192-4202CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
      Machine learning (ML) methods have become powerful, predictive tools in a wide range of applications, such as facial recognition and autonomous vehicles. In the sciences, computational chemists and physicists have been using ML for the prediction of phys. phenomena, such as atomistic potential energy surfaces and reaction pathways. Transferable ML potentials, such as ANI-1x, have been developed with the goal of accurately simulating org. mols. contg. the chem. elements H, C, N, and O. Here, we provide an extension of the ANI-1x model. The new model, dubbed ANI-2x, is trained to three addnl. chem. elements: S, F, and Cl. Addnl., ANI-2x underwent torsional refinement training to better predict mol. torsion profiles. These new features open a wide range of new applications within org. chem. and drug development. These seven elements (H, C, N, O, F, Cl, and S) make up ∼ 90% of drug-like mols. To show that these addns. do not sacrifice accuracy, we have tested this model across a range of org. mols. and applications, including the COMP6 benchmark, dihedral rotations, conformer scoring, and nonbonded interactions. ANI-2x is shown to accurately predict mol. energies compared to d. functional theory with a ~ 106 factor speedup and a negligible slowdown compared to ANI-1x and shows subchem. accuracy across most of the COMP6 benchmark. The resulting model is a valuable tool for drug development which can potentially replace both quantum calcns. and classical force fields for a myriad of applications.
    28. 28
      Qiao, Z.; Welborn, M.; Anandkumar, A.; Manby, F. R.; Miller, T. F. OrbNet: Deep learning for quantum chemistry using symmetry-adapted atomic-orbital features. J. Chem. Phys. 2020, 153, 124111,  DOI: 10.1063/5.0021955
      28
      OrbNet: Deep learning for quantum chemistry using symmetry-adapted atomic-orbital features
      Qiao, Zhuoran; Welborn, Matthew; Anandkumar, Animashree; Manby, Frederick R.; Miller, Thomas F.
      Journal of Chemical Physics (2020), 153 (12), 124111CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
      We introduce a machine learning method in which energy solns. from the Schroedinger equation are predicted using symmetry adapted AO features and a graph neural-network architecture. OrbNet is shown to outperform existing methods in terms of learning efficiency and transferability for the prediction of d. functional theory results while employing low-cost features that are obtained from semi-empirical electronic structure calcns. For applications to datasets of drug-like mols., including QM7b-T, QM9, GDB-13-T, DrugBank, and the conformer benchmark dataset of Folmsbee and Hutchison [Int. J. Quantum Chem. (published online) (2020)], OrbNet predicts energies within chem. accuracy of d. functional theory at a computational cost that is 1000-fold or more reduced. (c) 2020 American Institute of Physics.
    29. 29
      Christensen, A. S.; Sirumalla, S. K.; Qiao, Z.; O’Connor, M. B.; Smith, D. G. A.; Ding, F.; Bygrave, P. J.; Anandkumar, A.; Welborn, M.; Manby, F. R.; Miller, T. F. OrbNet Denali: A machine learning potential for biological and organic chemistry with semi-empirical cost and DFT accuracy. J. Chem. Phys. 2021, 155, 204103,  DOI: 10.1063/5.0061990
      29
      OrbNet Denali: A machine learning potential for biological and organic chemistry with semi-empirical cost and DFT accuracy
      Christensen, Anders S.; Sirumalla, Sai Krishna; Qiao, Zhuoran; O'Connor, Michael B.; Smith, Daniel G. A.; Ding, Feizhi; Bygrave, Peter J.; Anandkumar, Animashree; Welborn, Matthew; Manby, Frederick R.; Miller, Thomas F.
      Journal of Chemical Physics (2021), 155 (20), 204103CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
      We present OrbNet Denali, a machine learning model for an electronic structure that is designed as a drop-in replacement for ground-state d. functional theory (DFT) energy calcns. The model is a message-passing graph neural network that uses symmetry-adapted AO features from a low-cost quantum calcn. to predict the energy of a mol. OrbNet Denali is trained on a vast dataset of 2.3 x 106 DFT calcns. on mols. and geometries. This dataset covers the most common elements in biochem. and org. chem. (H, Li, B, C, N, O, F, Na, Mg, Si, P, S, Cl, K, Ca, Br, and I) and charged mols. OrbNet Denali is demonstrated on several well-established benchmark datasets, and we find that it provides accuracy that is on par with modern DFT methods while offering a speedup of up to three orders of magnitude. For the GMTKN55 benchmark set, OrbNet Denali achieves WTMAD-1 and WTMAD-2 scores of 7.19 and 9.84, on par with modern DFT functionals. For several GMTKN55 subsets, which contain chem. problems that are not present in the training set, OrbNet Denali produces a mean abs. error comparable to those of DFT methods. For the Hutchison conformer benchmark set, OrbNet Denali has a median correlation coeff. of R2 = 0.90 compared to the ref. DLPNO-CCSD(T) calcn. and R2 = 0.97 compared to the method used to generate the training data (ωB97X-D3/def2-TZVP), exceeding the performance of any other method with a similar cost. Similarly, the model reaches chem. accuracy for non-covalent interactions in the S66x10 dataset. For torsional profiles, OrbNet Denali reproduces the torsion profiles of ωB97X-D3/def2-TZVP with an av. mean abs. error of 0.12 kcal/mol for the potential energy surfaces of the diverse fragments in the TorsionNet500 dataset. (c) 2021 American Institute of Physics.
    30. 30
      Nakata, M.; Shimazaki, T. PubChemQC Project: A Large-Scale First-Principles Electronic Structure Database for Data-Driven Chemistry. J. Chem. Inf. Model. 2017, 57, 1300– 1308,  DOI: 10.1021/acs.jcim.7b00083
      30
      PubChemQC Project: A Large-Scale First-Principles Electronic Structure Database for Data-Driven Chemistry
      Nakata, Maho; Shimazaki, Tomomi
      Journal of Chemical Information and Modeling (2017), 57 (6), 1300-1308CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      Large-scale mol. databases play an essential role in the investigation of various subjects such as the development of org. materials, in-silico drug designs, and data-driven studies with machine learning, among others. We developed a large-scale quantum chem. database based on the first-principles method without performing any expt. Our database currently contains three million mol. electronic structures based on the d. functional theory method at the B3LYP/6-31G* level, and we successively calcd. 10 low-lying excited states of over two million mols. by the time-dependent DFT method with the 6-31+G* basis set. To select the mols. calcd. in our project, we mainly referred to the PubChem project, and it was used as a source of the mol. structures in short strings using the InChI and the SMILES representations. Accordingly, we named our quantum chem. database project as "PubChemQC" (http://pubchemqc.riken.jp/) and placed it in the public domain. In this paper, we showed the fundamental features of the PubChemQC database and dis- cussed the techniques used to construct the dataset for large-scale quantum chem. calcns. We also presented a machine-learning approach to predict the electronic structure of mols. as an example to demonstrate the suitability of the large-scale quantum chem. database.
    31. 31
      Smith, D. G. A.; Altarawy, D.; Burns, L. A.; Welborn, M.; Naden, L. N.; Ward, L.; Ellis, S.; Pritchard, B. P.; Crawford, T. D. The MolSSI QCArchive project: An open-source platform to compute, organize, and share quantum chemistry data. Wiley Interdiscip. Rev. Comput. Mol. Sci. 2020, 11, e1491  DOI: 10.1002/wcms.1491
      There is no corresponding record for this reference.
    32. 32
      Lim, V. T.; Hahn, D. F.; Tresadern, G.; Bayly, C. I.; Mobley, D. L. Benchmark assessment of molecular geometries and energies from small molecule force fields [version 1; peer review: 2 approved]. F1000Research 2020, 9, 1390,  DOI: 10.12688/f1000research.27141.1
      32
      Benchmark assessment of molecular geometries and energies from small molecule force fields
      Lim, Victoria T.; Hahn, David F.; Tresadern, Gary; Bayly, Christopher I.; Mobley, David L.
      F1000Research (2020), 9 (), 1390CODEN: FRESJL; ISSN:2046-1402. (F1000 Research Ltd.)
      Background: Force fields are used in a wide variety of contexts for classical mol. simulation, including studies on protein-ligand binding, membrane permeation, and thermophys. property prediction. The quality of these studies relies on the quality of the force fields used to represent the systems. Methods: Focusing on small mols. of fewer than 50 heavy atoms, our aim in this work is to compare nine force fields: GAFF, GAFF2, MMFF94, MMFF94S, OPLS3e, SMIRNOFF99Frosst, and the Open Force Field Parsley, versions 1.0, 1.1, and 1.2. On a dataset comprising 22,675 mol. structures of 3,271 mols., we analyzed force field-optimized geometries and conformer energies compared to ref. quantum mech. (QM) data. Results: We show that while OPLS3e performs best, the latest Open Force Field Parsley release is approaching a comparable level of accuracy in reproducing QM geometries and energetics for this set of mols. Meanwhile, the performance of established force fields such as MMFF94S and GAFF2 is generally somewhat worse. We also fiend that the series of recent Open Force Field versions provide significant increases in accuracy. Conclusions: This study provides an extensive test of the performance of different mol. mechanics force fields on a diverse mol. set, and highlights two (OPLS3e and OpenFF 1.2) that perform better than the others tested on the present comparison. Our mol. set and results are available for other researchers to use in testing.
    33. 33
      Gražulis, S.; Merkys, A.; Vaitkus, A.; Chateigner, D.; Lutterotti, L.; Moeck, P.; Quiros, M.; Downs, R. T.; Kaminsky, W.; Bail, A. L. Materials Informatics: Methods, Tools and Applications; Isayev, O., Tropsha, A., Curtarolo, S., Eds.; Wiley, 2019; Chapter 1, pp 1– 39.
      There is no corresponding record for this reference.
    34. 34
      Chai, J.-D.; Head-Gordon, M. Systematic optimization of long-range corrected hybrid density functionals. J. Chem. Phys. 2008, 128, 084106,  DOI: 10.1063/1.2834918
      34
      Systematic optimization of long-range corrected hybrid density functionals
      Chai, Jeng-Da; Head-Gordon, Martin
      Journal of Chemical Physics (2008), 128 (8), 084106/1-084106/15CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
      A general scheme for systematically modeling long-range cor. (LC) hybrid d. functionals is proposed. Our resulting two LC hybrid functionals are shown to be accurate in thermochem., kinetics, and noncovalent interactions, when compared with common hybrid d. functionals. The qual. failures of the commonly used hybrid d. functionals in some "difficult problems," such as dissocn. of sym. radical cations and long-range charge-transfer excitations, are significantly reduced by the present LC hybrid d. functionals. (c) 2008 American Institute of Physics.
    35. 35
      Weigend, F.; Ahlrichs, R. Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy. Phys. Chem. Chem. Phys. 2005, 7, 3297,  DOI: 10.1039/b508541a
      35
      Balanced basis sets of split valence, triple zeta valence and quadruple zeta valence quality for H to Rn: Design and assessment of accuracy
      Weigend, Florian; Ahlrichs, Reinhart
      Physical Chemistry Chemical Physics (2005), 7 (18), 3297-3305CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
      Gaussian basis sets of quadruple zeta valence quality for Rb-Rn are presented, as well as bases of split valence and triple zeta valence quality for H-Rn. The latter were obtained by (partly) modifying bases developed previously. A large set of more than 300 mols. representing (nearly) all elements-except lanthanides-in their common oxidn. states was used to assess the quality of the bases all across the periodic table. Quantities investigated were atomization energies, dipole moments and structure parameters for Hartree-Fock, d. functional theory and correlated methods, for which we had chosen Moller-Plesset perturbation theory as an example. Finally recommendations are given which type of basis set is used best for a certain level of theory and a desired quality of results.
    36. 36
      Weigend, F. Accurate Coulomb-fitting basis sets for H to Rn. Phys. Chem. Chem. Phys. 2006, 8, 1057,  DOI: 10.1039/b515623h
      36
      Accurate Coulomb-fitting basis sets for H to Rn
      Weigend, Florian
      Physical Chemistry Chemical Physics (2006), 8 (9), 1057-1065CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
      A series of auxiliary basis sets to fit Coulomb potentials for the elements H to Rn (except lanthanides) is presented. For each element only one auxiliary basis set is needed to approx. Coulomb energies in conjunction with orbital basis sets of split valence, triple zeta valence and quadruple zeta valence quality with errors of typically below ca. 0.15 kJ mol-1 per atom; this was demonstrated in conjunction with the recently developed orbital basis sets of types def2-SV(P), def2-TZVP and def2-QZVPP for a large set of small mols. representing (nearly) each element in all of its common oxidn. states. These auxiliary bases are slightly more than three times larger than orbital bases of split valence quality. Compared to non-approximated treatments, computation times for the Coulomb part are reduced by a factor of ca. 8 for def2-SV(P) orbital bases, ca. 25 for def2-TZVP and ca. 100 for def2-QZVPP orbital bases.
    37. 37
      Sterling, T.; Irwin, J. J. ZINC 15 – Ligand Discovery for Everyone. J. Chem. Inf. Model. 2015, 55, 2324– 2337,  DOI: 10.1021/acs.jcim.5b00559
      37
      ZINC 15 - Ligand Discovery for Everyone
      Sterling, Teague; Irwin, John J.
      Journal of Chemical Information and Modeling (2015), 55 (11), 2324-2337CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      Many questions about the biol. activity and availability of small mols. remain inaccessible to investigators who could most benefit from their answers. To narrow the gap between chemoinformatics and biol., we have developed a suite of ligand annotation, purchasability, target, and biol. assocn. tools, incorporated into ZINC and meant for investigators who are not computer specialists. The new version contains over 120 million purchasable "drug-like" compds. - effectively all org. mols. that are for sale - a quarter of which are available for immediate delivery. ZINC connects purchasable compds. to high-value ones such as metabolites, drugs, natural products, and annotated compds. from the literature. Compds. may be accessed by the genes for which they are annotated as well as the major and minor target classes to which those genes belong. It offers new anal. tools that are easy for nonspecialists yet with few limitations for experts. ZINC retains its original 3D roots - all mols. are available in biol. relevant, ready-to-dock formats. ZINC is freely available at http://zinc15.docking.org.
    38. 38
      Yoshikawa, N.; Hutchison, G. R. Fast, efficient fragment-based coordinate generation for Open Babel. J. Cheminf. 2019, 11, 49,  DOI: 10.1186/s13321-019-0372-5
      There is no corresponding record for this reference.
    39. 39
      O’Boyle, N. M.; Morley, C.; Hutchison, G. R. Pybel: a Python wrapper for the OpenBabel cheminformatics toolkit. Chem. Cent. J. 2008, 2, 5,  DOI: 10.1186/1752-153x-2-5
      39
      Pybel: a Python wrapper for the OpenBabel cheminformatics toolkit
      O'Boyle Noel M; Morley Chris; Hutchison Geoffrey R
      Chemistry Central journal (2008), 2 (), 5 ISSN:.
      BACKGROUND: Scripting languages such as Python are ideally suited to common programming tasks in cheminformatics such as data analysis and parsing information from files. However, for reasons of efficiency, cheminformatics toolkits such as the OpenBabel toolkit are often implemented in compiled languages such as C++. We describe Pybel, a Python module that provides access to the OpenBabel toolkit. RESULTS: Pybel wraps the direct toolkit bindings to simplify common tasks such as reading and writing molecular files and calculating fingerprints. Extensive use is made of Python iterators to simplify loops such as that over all the molecules in a file. A Pybel Molecule can be easily interconverted to an OpenBabel OBMol to access those methods or attributes not wrapped by Pybel. CONCLUSION: Pybel allows cheminformaticians to rapidly develop Python scripts that manipulate chemical information. It is open source, available cross-platform, and offers the power of the OpenBabel toolkit to Python programmers.
    40. 40
      O’Boyle, N. M.; Banck, M.; James, C. A.; Morley, C.; Vandermeersch, T.; Hutchison, G. R. Open Babel: An open chemical toolbox. J. Cheminf. 2011, 3, 33,  DOI: 10.1186/1758-2946-3-33
      40
      Open Babel: an open chemical toolbox
      O'Boyle, Noel M.; Banck, Michael; James, Craig A.; Morley, Chris; Vandermeersch, Tim; Hutchison, Geoffrey R.
      Journal of Cheminformatics (2011), 3 (), 33CODEN: JCOHB3; ISSN:1758-2946. (Chemistry Central Ltd.)
      Background: A frequent problem in computational modeling is the interconversion of chem. structures between different formats. While std. interchange formats exist (for example, Chem. Markup Language) and de facto stds. have arisen (for example, SMILES format), the need to interconvert formats is a continuing problem due to the multitude of different application areas for chem. data, differences in the data stored by different formats (0D vs. 3D, for example), and competition between software along with a lack of vendor-neutral formats. Results: We discuss, for the first time, Open Babel, an open-source chem. toolbox that speaks the many languages of chem. data. Open Babel version 2.3 interconverts over 110 formats. The need to represent such a wide variety of chem. and mol. data requires a library that implements a wide range of cheminformatics algorithms, from partial charge assignment and aromaticity detection, to bond order perception and canonicalization. We detail the implementation of Open Babel, describe key advances in the 2.3 release, and outline a variety of uses both in terms of software products and scientific research, including applications far beyond simple format interconversion. Conclusions: Open Babel presents a soln. to the proliferation of multiple chem. file formats. In addn., it provides a variety of useful utilities from conformer searching and 2D depiction, to filtering, batch conversion, and substructure and similarity searching. For developers, it can be used as a programming library to handle chem. data in areas such as org. chem., drug design, materials science, and computational chem. It is freely available under an open-source license.
    41. 41
      Wishart, D. S.; Feunang, Y. D.; Guo, A. C.; Lo, E. J.; Marcu, A.; Grant, J. R.; Sajed, T.; Johnson, D.; Li, C.; Sayeeda, Z. DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic Acids Res. 2018, 46, D1074– D1082,  DOI: 10.1093/nar/gkx1037
      41
      DrugBank 5.0: a major update to the DrugBank database for 2018
      Wishart, David S.; Feunang, Yannick D.; Guo, An C.; Lo, Elvis J.; Marcu, Ana; Grant, Jason R.; Sajed, Tanvir; Johnson, Daniel; Li, Carin; Sayeeda, Zinat; Assempour, Nazanin; Iynkkaran, Ithayavani; Liu, Yifeng; Maciejewski, Adam; Gale, Nicola; Wilson, Alex; Chin, Lucy; Cummings, Ryan; Le, Diana; Pon, Allison; Knox, Craig; Wilson, Michael
      Nucleic Acids Research (2018), 46 (D1), D1074-D1082CODEN: NARHAD; ISSN:1362-4962. (Oxford University Press)
      DrugBank is a web-enabled database contg. comprehensivemol. information about drugs, their mechanisms, their interactions and their targets. First described in 2006, Drug- Bank has continued to evolve over the past 12 years in response to marked improvements to web stds. and changing needs for drug research and development. This year's update, DrugBank 5.0, represents the most significant upgrade to the database in more than 10 years. In many cases, existing data content has grown by 100% or more over the last update. For instance, the total no. of investigational drugs in the database has grown by almost 300%, the no. of drug-drug interactions has grown by nearly 600% and the no. of SNP-assocd. drug effects has grown more than 3000%. Significant improvements have been made to the quantity, quality and consistency of drug indications, drug binding data as well as drug-drug and drug-food interactions. A great deal of brand new data have also been added to DrugBank 5.0. This includes information on the influence of hundreds of drugs on metabolite levels (pharmacometabolomics), gene expression levels (pharmacotranscriptomics) and protein expression levels (pharmacoproteomics). New data have also been added on the status of hundreds of newdrug clin. trials and existing drug repurposing trials. Many other important improvements in the content, interface and performance of the DrugBank website have been made and these should greatly enhance its ease of use, utility and potential applications in many areas of pharmacol. research, pharmaceutical science and drug education.
    42. 42
      Pracht, P.; Bohle, F.; Grimme, S. Automated Exploration of the low-energy Chemical Space with fast Quantum Chemical Methods. Phys. Chem. Chem. Phys. 2020, 22, 7169– 7192,  DOI: 10.1039/C9CP06869D
      42
      Automated exploration of the low-energy chemical space with fast quantum chemical methods
      Pracht, Philipp; Bohle, Fabian; Grimme, Stefan
      Physical Chemistry Chemical Physics (2020), 22 (14), 7169-7192CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
      We propose and discuss an efficient scheme for the in silico sampling for parts of the mol. chem. space by semiempirical tight-binding methods combined with a meta-dynamics driven search algorithm. The focus of this work is set on the generation of proper thermodn. ensembles at a quantum chem. level for conformers, but similar procedures for protonation states, tautomerism and non-covalent complex geometries are also discussed. The conformational ensembles consisting of all significantly populated min. energy structures normally form the basis of further, mostly DFT computational work, such as the calcn. of spectra or macroscopic properties. By using basic quantum chem. methods, electronic effects or possible bond breaking/formation are accounted for and a very reasonable initial energetic ranking of the candidate structures is obtained. Due to the huge computational speedup gained by the fast low-cost quantum chem. methods, overall short computation times even for systems with hundreds of atoms (typically drug-sized mols.) are achieved. Furthermore, specialized applications, such as sampling with implicit solvation models or constrained conformational sampling for transition-states, metal-, surface-, or noncovalently bound complexes are discussed, opening many possible applications in modern computational chem. and drug discovery. The procedures have been implemented in a freely available computer code called CREST, that makes use of the fast and reliable GFNn-xTB methods.
    43. 43
      Grimme, S. Exploration of Chemical Compound, Conformer, and Reaction Space with Meta-Dynamics Simulations Based on Tight-Binding Quantum Chemical Calculations. J. Chem. Theory Comput. 2019, 15, 2847– 2862,  DOI: 10.1021/acs.jctc.9b00143
      43
      Exploration of Chemical Compound, Conformer, and Reaction Space with Meta-Dynamics Simulations Based on Tight-Binding Quantum Chemical Calculations
      Grimme, Stefan
      Journal of Chemical Theory and Computation (2019), 15 (5), 2847-2862CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
      The semiempirical tight-binding based quantum chem. method GFN2-xTB is used in the framework of meta-dynamics (MTD) to globally explore chem. compd., conformer, and reaction space. The biasing potential given as a sum of Gaussian functions is expressed with the root-mean-square-deviation (RMSD) in Cartesian space as a metric for the collective variables. This choice makes the approach robust and generally applicable to three common problems (i.e., conformer search, chem. reaction space exploration in a virtual nanoreactor, and for guessing reaction paths). Because of the inherent locality of the at. RMSD, functional group or fragment selective treatments are possible facilitating the investigation of catalytic processes where, for example, only the substrate is thermally activated. Due to the approx. character of the GFN2-xTB method, the resulting structure ensembles require further refinement with more sophisticated, for example, d. functional or wave function theory methods. However, the approach is extremely efficient running routinely on common laptop computers in minutes to hours of computation time even for realistically sized mols. with a few hundred atoms. Furthermore, the underlying potential energy surface for mols. contg. almost all elements (Z = 1-86) is globally consistent including the covalent dissocn. process and electronically complicated situations in, for example, transition metal systems. As examples, thermal decompn., ethyne oligomerization, the oxidn. of hydrocarbons (by oxygen and a P 450 enzyme model), a Miller-Urey model system, a thermally forbidden dimerization, and a multistep intramol. cyclization reaction are shown. For typical conformational search problems of org. drug mols., the new MTD(RMSD) algorithm yields lower energy structures and more complete conformer ensembles at reduced computational effort compared with its already well performing predecessor.
    44. 44
      Chan, L.; Morris, G. M.; Hutchison, G. R. Understanding Conformational Entropy in Small Molecules. J. Chem. Theory Comput. 2021, 17, 2099– 2106,  DOI: 10.1021/acs.jctc.0c01213
      44
      Understanding Conformational Entropy in Small Molecules
      Chan, Lucian; Morris, Garrett M.; Hutchison, Geoffrey R.
      Journal of Chemical Theory and Computation (2021), 17 (4), 2099-2106CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
      The calcn. of the entropy of flexible mols. can be challenging, since the no. of possible conformers can grow exponentially with mol. size and many low-energy conformers may be thermally accessible. Different methods have been proposed to approx. the contribution of conformational entropy to the mol. std. entropy, including performing thermochem. calcns. with all possible stable conformations and developing empirical corrections from exptl. data. We have performed conformer sampling on over 120,000 small mols. generating some 12 million conformers, to develop models to predict conformational entropy across a wide range of mols. Using insight into the nature of conformational disorder, our cross-validated phys. motivated statistical model gives a mean abs. error of ∼ 4.8 J/mol·K or under 0.4 kcal/mol at 300 K. Beyond predicting mol. entropies and free energies, the model implies a high degree of correlation between torsions in most mols., often assumed to be independent. While individual dihedral rotations may have low energetic barriers, the shape and chem. functionality of most mols. necessarily correlate their torsional degrees of freedom and hence restrict the no. of low-energy conformations immensely. Our simple models capture these correlations and advance our understanding of small mol. conformational entropy.
    45. 45
      Landrum, G. RDKit: Open-Source Cheminformatics. Available at http://www.rdkit.org, 2020; http://www.rdkit.org (accesses Oct 1, 2022).
      There is no corresponding record for this reference.
    46. 46
      https://smarts.plus/.
      There is no corresponding record for this reference.
    47. 47
      Schomburg, K.; Ehrlich, H.-C.; Stierand, K.; Rarey, M. From Structure Diagrams to Visual Chemical Patterns. J. Chem. Inf. Model. 2010, 50, 1529– 1535,  DOI: 10.1021/ci100209a
      47
      From Structure Diagrams to Visual Chemical Patterns
      Schomburg, Karen; Ehrlich, Hans-Christian; Stierand, Katrin; Rarey, Matthias
      Journal of Chemical Information and Modeling (2010), 50 (9), 1529-1535CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The intuitive way of chemists to communicate mols. is via two-dimensional structure diagrams. The straightforward visual representations are mostly preferred to the often complicated systematic chem. names. For chem. patterns, however, no comparable visualization stds. have evolved so far. Chem. patterns denoting descriptions of chem. features are needed whenever a set of mols. is filtered for certain properties. The currently available representations are constrained to linear mol. pattern languages which are hardly human readable and therefore keep chemists without computational background from systematically formulating patterns. Therefore, we introduce a new visualization concept for chem. patterns. The common std. concept of structure diagrams is extended to account for property descriptions and logic combinations of chem. features in patterns. As a first application of the new concept, we developed the SMARTSviewer, a tool that converts chem. patterns encoded in SMARTS strings to a visual representation. The graphic pattern depiction provides an overview of the specified chem. features, variations, and similarities without needing to decode the often cryptic linear expressions. Taking recent chem. publications from various fields, we demonstrate the wide application range of a graphical chem. pattern language.
    48. 48
      Neese, F.; Wennmohs, F.; Becker, U.; Riplinger, C. The ORCA quantum chemistry program package. J. Chem. Phys. 2020, 152, 224108,  DOI: 10.1063/5.0004608
      48
      The ORCA quantum chemistry program package
      Neese, Frank; Wennmohs, Frank; Becker, Ute; Riplinger, Christoph
      Journal of Chemical Physics (2020), 152 (22), 224108CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
      In this contribution to the special software-centered issue, the ORCA program package is described. We start with a short historical perspective of how the project began and go on to discuss its current feature set. ORCA has grown into a rather comprehensive general-purpose package for theor. research in all areas of chem. and many neighboring disciplines such as materials sciences and biochem. ORCA features d. functional theory, a range of wavefunction based correlation methods, semi-empirical methods, and even force-field methods. A range of solvation and embedding models is featured as well as a complete intrinsic to ORCA quantum mechanics/mol. mechanics engine. A specialty of ORCA always has been a focus on transition metals and spectroscopy as well as a focus on applicability of the implemented methods to "real-life" chem. applications involving systems with a few hundred atoms. In addn. to being efficient, user friendly, and, to the largest extent possible, platform independent, ORCA features a no. of methods that are either unique to ORCA or have been first implemented in the course of the ORCA development. Next to a range of spectroscopic and magnetic properties, the linear- or low-order single- and multi-ref. local correlation methods based on pair natural orbitals (domain based local pair natural orbital methods) should be mentioned here. Consequently, ORCA is a widely used program in various areas of chem. and spectroscopy with a current user base of over 22 000 registered users in academic research and in industry. (c) 2020 American Institute of Physics.
    49. 49
      Lin, J. B.; Jin, Y.; Lopez, S. A.; Druckerman, N.; Wheeler, S. E.; Houk, K. N. Torsional Barriers to Rotation and Planarization in Heterocyclic Oligomers of Value in Organic Electronics. J. Chem. Theory Comput. 2017, 13, 5624– 5638,  DOI: 10.1021/acs.jctc.7b00709
      49
      Torsional Barriers to Rotation and Planarization in Heterocyclic Oligomers of Value in Organic Electronics
      Lin, Janice B.; Jin, Yu; Lopez, Steven A.; Druckerman, Nathaniel; Wheeler, Steven E.; Houk, K. N.
      Journal of Chemical Theory and Computation (2017), 13 (11), 5624-5638CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
      In order to understand the conformational behavior of org. components in org. electronic devices, we have computed the torsional potentials for a library of thiophene-based heterodimers. The accuracy and efficiencies of computational methods for these org. materials were benchmarked for 11 common d. functionals with three Pople basis sets against a Focal Point Anal. (FPA) on a model oligothiophene 2,5-bis(3-tetradecylthiophen-2-yl)thieno[3,2-b]-thiophene (BTTT) system. This study establishes a set of general trends in regards to conformational preferences, as well as planarization and rotational barriers for a library comprised of common fragments found in org. materials. These gas phase structures are compared to exptl. crystal structures to det. the effect of crystal packing on geometry. Finally, we analyze the structure of hole-transporting material DERDTS-TBDT and design a new oligomer likely to be planar in the solid state.
    50. 50
      Perkins, M. A.; Cline, L. M.; Tschumper, G. S. Torsional Profiles of Thiophene and Furan Oligomers: Probing the Effects of Heterogeneity and Chain Length. J. Phys. Chem. A 2021, 125, 6228– 6237,  DOI: 10.1021/acs.jpca.1c04714
      50
      Torsional Profiles of Thiophene and Furan Oligomers: Probing the Effects of Heterogeneity and Chain Length
      Perkins, Morgan A.; Cline, Laura M.; Tschumper, Gregory S.
      Journal of Physical Chemistry A (2021), 125 (28), 6228-6237CODEN: JPCAFH; ISSN:1089-5639. (American Chemical Society)
      A systematic anal. of the torsional profiles of 55 unique oligomers composed of two to four thiophene and/or furan rings (n = 2 to 4) has been conducted using three d. functional theory (DFT) methods along with MP2 and three different coupled-cluster methods. Two planar or quasi-planar min. were identified for each n = 2 oligomer system. In every case, the torsional angle (τ) between the heteroatoms about the carbon-carbon bond connecting the two rings is at or near 180° for the global min. and 0° for the local min., referred to as anti and syn conformations, resp. These oligomers have rotational barrier heights ranging from ca. 2 kcal mol-1 for 2,2'-bithiophene to 4 kcal mol-1 for 2,2'-bifuran, based on electronic energies computed near the CCSD(T) complete basis set (CBS) limit. The corresponding rotational barrier for the heterogeneous 2-(2-thienyl)furan counterpart falls approx. halfway between those values. The energy differences between the min. are approx. 2 and 0.4 kcal mol-1 for the homogeneous 2,2'-bifuran and 2,2'-bithiophene, resp., whereas the energy difference between the planar local and global min. (at τ = 0 and 180°, resp.) is only 0.3 kcal mol-1 for 2-(2-thienyl)furan. Extending these three oligomers by adding one or two addnl. thiophene and/or furan rings resulted in only minor changes to the torsional profiles when rotating around the same carbon-carbon bond as the two-ring profiles. Relative energy differences between the syn and anti conformations were changed by no more than 0.4 kcal mol-1 for the corresponding n = 3 and 4 oligomers, while the rotational barrier height increased by no more than 0.8 kcal mol-1.
    51. 51
      Johansson, M. P.; Olsen, J. Torsional Barriers and Equilibrium Angle of Biphenyl: Reconciling Theory with Experiment. J. Chem. Theory Comput. 2008, 4, 1460– 1471,  DOI: 10.1021/ct800182e
      51
      Torsional Barriers and Equilibrium Angle of Biphenyl: Reconciling Theory with Experiment
      Johansson, Mikael P.; Olsen, Jeppe
      Journal of Chemical Theory and Computation (2008), 4 (9), 1460-1471CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
      The barriers of internal rotation of the two Ph groups in biphenyl are investigated using a combination of coupled cluster and d. functional theory. The exptl. barriers are for the first time accurately reproduced; our best ests. of the barriers are 8.0 and 8.3 kJ/mol around the planar and perpendicular conformations, resp. The use of flexible basis sets of at least augmented quadruple-ζ quality is shown to be a crucial prerequisite. Further, to finally reconcile theory with expt., extrapolations of both the basis set toward the basis set limit and electron correlation toward the full configuration-interaction limit are necessary. The min. of the torsional angle is significantly increased by free energy corrections, which are needed to reach an agreement with expt. The d. functional B3LYP approach is found to perform well compared with the highest level ab initio results.
    52. 52
      Nam, S.; Cho, E.; Sim, E.; Burke, K. Explaining and Fixing DFT Failures for Torsional Barriers. J. Phys. Chem. Lett. 2021, 12, 2796– 2804,  DOI: 10.1021/acs.jpclett.1c00426
      52
      Explaining and Fixing DFT Failures for Torsional Barriers
      Nam, Seungsoo; Cho, Eunbyol; Sim, Eunji; Burke, Kieron
      Journal of Physical Chemistry Letters (2021), 12 (11), 2796-2804CODEN: JPCLCD; ISSN:1948-7185. (American Chemical Society)
      Most torsional barriers are predicted with high accuracies (about 1 kJ/mol) by std. semilocal functionals, but a small subset was found to have much larger errors. We created a database of almost 300 carbon-carbon torsional barriers, including 12 poorly behaved barriers, that stem from the Y=C-X group, where Y is O or S and X is a halide. Functionals with enhanced exchange mixing (about 50%) worked well for all barriers. We found that poor actors have delocalization errors caused by hyperconjugation. These problematic calcns. are d.-sensitive (i.e., DFT predictions change noticeably with the d.), and using HF densities (HF-DFT) fixes these issues. For example, conventional B3LYP performs as accurately as exchange-enhanced functionals if the HF d. is used. For long-chain conjugated mols., HF-DFT can be much better than exchange-enhanced functionals. We suggest that HF-PBE0 has the best overall performance.
    53. 53
      Jackson, N. E.; Savoie, B. M.; Kohlstedt, K. L.; Olvera de la Cruz, M.; Schatz, G. C.; Chen, L. X.; Ratner, M. A. Controlling Conformations of Conjugated Polymers and Small Molecules: The Role of Nonbonding Interactions. J. Am. Chem. Soc. 2013, 135, 10475– 10483,  DOI: 10.1021/ja403667s
      53
      Controlling Conformations of Conjugated Polymers and Small Molecules: The Role of Nonbonding Interactions
      Jackson, Nicholas E.; Savoie, Brett M.; Kohlstedt, Kevin L.; Olvera de la Cruz, Monica; Schatz, George C.; Chen, Lin X.; Ratner, Mark A.
      Journal of the American Chemical Society (2013), 135 (28), 10475-10483CODEN: JACSAT; ISSN:0002-7863. (American Chemical Society)
      The chem. variety present in the org. electronics literature has motivated us to investigate potential nonbonding interactions often incorporated into conformational "locking" schemes. We examine a variety of potential interactions, including oxygen-sulfur, nitrogen-sulfur, and fluorine-sulfur, using accurate quantum-chem. wave function methods and noncovalent interaction (NCI) anal. on a selection of high-performing conjugated polymers and small mols. found in the literature. In addn., we evaluate a set of nonbonding interactions occurring between various heterocyclic and pendant atoms taken from a group of representative π-conjugated mols. Together with our survey and set of interactions, it is detd. that while many nonbonding interactions possess weak binding capabilities, nontraditional hydrogen-bonding interactions, oxygen-hydrogen (CH···O) and nitrogen-hydrogen (CH···N), are alone in inducing conformational control and enhanced planarity along a polymer or small mol. backbone at room temp.
    54. 54
      Greenwell, C.; Beran, G. J. O. Inaccurate Conformational Energies Still Hinder Crystal Structure Prediction in Flexible Organic Molecules. Cryst. Growth Des. 2020, 20, 4875– 4881,  DOI: 10.1021/acs.cgd.0c00676
      54
      Inaccurate Conformational Energies Still Hinder Crystal Structure Prediction in Flexible Organic Molecules
      Greenwell, Chandler; Beran, Gregory J. O.
      Crystal Growth & Design (2020), 20 (8), 4875-4881CODEN: CGDEFU; ISSN:1528-7483. (American Chemical Society)
      Crystal structure prediction driven by d. functional theory has become an increasingly useful tool for the pharmaceutical industry and others interested in understanding and controlling org. mol. crystal packing. However, delocalization error in widely used d. functionals leads to problematic conformational energies that can cause incorrect predictions of polymorph stabilities. In 5 examples ranging from small mols. to the polymorphically challenging pharmaceuticals axitinib and galunisertib, inexpensively correcting the intramol. conformational energies with higher-level electronic structure methods leads to polymorph stability predictions that agree far better with expt. This approach also provides a valuable diagnostic for when skepticism about predicted polymorph stabilities is warranted. Commonly used d. functionals have difficulty ranking certain types of conformational polymorph structures correctly. Correcting the intramol. conformational energies with higher-level quantum chem. methods can improve the accuracy of crystal structure prediction stability rankings considerably.
    55. 55
      Smith, J. S.; Zubatyuk, R.; Nebgen, B.; Lubbers, N.; Barros, K.; Roitberg, A. E.; Isayev, O.; Tretiak, S. The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules. Sci. Data 2020, 7, 134,  DOI: 10.1038/s41597-020-0473-z
      55
      The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules
      Smith, Justin S.; Zubatyuk, Roman; Nebgen, Benjamin; Lubbers, Nicholas; Barros, Kipton; Roitberg, Adrian E.; Isayev, Olexandr; Tretiak, Sergei
      Scientific Data (2020), 7 (1), 134CODEN: SDCABS; ISSN:2052-4463. (Nature Research)
      Abstr.: Maximum diversification of data is a central theme in building generalized and accurate machine learning (ML) models. In chem., ML has been used to develop models for predicting mol. properties, for example quantum mechanics (QM) calcd. potential energy surfaces and at. charge models. The ANI-1x and ANI-1ccx ML-based general-purpose potentials for org. mols. were developed through active learning; an automated data diversification process. Here, we describe the ANI-1x and ANI-1ccx data sets. To demonstrate data diversity, we visualize it with a dimensionality redn. scheme, and contrast against existing data sets. The ANI-1x data set contains multiple QM properties from 5 M d. functional theory calcns., while the ANI-1ccx data set contains 500 k data points obtained with an accurate CCSD(T)/CBS extrapolation. Approx. 14 million CPU core-hours were expended to generate this data. Multiple QM calcd. properties for the chem. elements C, H, N, and O are provided: energies, at. forces, multipole moments, at. charges, etc. We provide this data to the community to aid research and development of ML models for chem.
    56. 56
      Axelrod, S.; Gomez-Bombarelli, R. GEOM, energy-annotated molecular conformations for property prediction and molecular generation. Sci. Data 2022, 9, 185,  DOI: 10.1038/s41597-022-01288-4
      56
      GEOM, energy-annotated molecular conformations for property prediction and molecular generation
      Axelrod, Simon; Gomez-Bombarelli, Rafael
      Scientific Data (2022), 9 (1), 185CODEN: SDCABS; ISSN:2052-4463. (Nature Portfolio)
      Machine learning (ML) outperforms traditional approaches in many mol. design tasks. ML models usually predict mol. properties from a 2D chem. graph or a single 3D structure, but neither of these representations accounts for the ensemble of 3D conformers that are accessible to a mol. Property prediction could be improved by using conformer ensembles as input, but there is no large-scale dataset that contains graphs annotated with accurate conformers and exptl. data. Here we use advanced sampling and semi-empirical d. functional theory (DFT) to generate 37 million mol. conformations for over 450,000 mols. The Geometric Ensemble Of Mols. (GEOM) dataset contains conformers for 133,000 species from QM9, and 317,000 species with exptl. data related to biophysics, physiol., and phys. chem. Ensembles of 1,511 species with BACE-1 inhibition data are also labeled with high-quality DFT free energies in an implicit water solvent, and 534 ensembles are further optimized with DFT. GEOM will assist in the development of models that predict properties from conformer ensembles, and generative models that sample 3D conformations.
    57. 57
      Isert, C.; Atz, K.; Jiménez-Luna, J.; Schneider, G. QMugs, quantum mechanical properties of drug-like molecules. Sci. Data 2022, 9, 273,  DOI: 10.1038/s41597-022-01390-7
      57
      QMugs, quantum mechanical properties of drug-like molecules
      Isert, Clemens; Atz, Kenneth; Jimenez-Luna, Jose; Schneider, Gisbert
      Scientific Data (2022), 9 (1), 273CODEN: SDCABS; ISSN:2052-4463. (Nature Portfolio)
      Machine learning approaches in drug discovery, as well as in other areas of the chem. sciences, benefit from curated datasets of phys. mol. properties. However, there currently is a lack of data collections featuring large bioactive mols. alongside first-principle quantum chem. information. The open-access QMugs (Quantum-Mech. Properties of Drug-like Mols.) dataset fills this void. The QMugs collection comprises quantum mech. properties of more than 665 k biol. and pharmacol. relevant mols. extd. from the ChEMBL database, totaling ∼2 M conformers. QMugs contains optimized mol. geometries and thermodn. data obtained via the semi-empirical method GFN2-xTB. Atomic and mol. properties are provided on both the GFN2-xTB and on the d.-functional levels of theory (DFT, ωB97X-D/def2-SVP). QMugs features mols. of significantly larger size than previously-reported collections and comprises their resp. quantum mech. wave functions, including DFT d. and orbital matrixes. This dataset is intended to facilitate the development of models that learn from mol. data on different levels of theory while also providing insight into the corresponding relationships between mol. structure and biol. activity.
    58. 58
      Eastman, P.; Behara, P. K.; Dotson, D. L.; Galvelis, R.; Herr, J. E.; Horton, J. T.; Mao, Y.; Chodera, J. D.; Pritchard, B. P.; Wang, Y.; Fabritiis, G. D.; Markland, T. E. SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials. Sci. Data 2023, 10, 11,  DOI: 10.1038/s41597-022-01882-6
      58
      SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine Learning Potentials
      Eastman, Peter; Behara, Pavan Kumar; Dotson, David L.; Galvelis, Raimondas; Herr, John E.; Horton, Josh T.; Mao, Yuezhi; Chodera, John D.; Pritchard, Benjamin P.; Wang, Yuanqing; De Fabritiis, Gianni; Markland, Thomas E.
      Scientific Data (2023), 10 (1), 11CODEN: SDCABS; ISSN:2052-4463. (Nature Portfolio)
      Machine learning potentials are an important tool for mol. simulation, but their development is held back by a shortage of high quality datasets to train them on. We describe the SPICE dataset, a new quantum chem. dataset for training potentials relevant to simulating drug-like small mols. interacting with proteins. It contains over 1.1 million conformations for a diverse set of small mols., dimers, dipeptides, and solvated amino acids. It includes 15 elements, charged and uncharged mols., and a wide range of covalent and non-covalent interactions. It provides both forces and energies calcd. at the ωB97M-D3(BJ)/def2-TZVPPD level of theory, along with other useful quantities such as multipole moments and bond orders. We train a set of machine learning potentials on it and demonstrate that they can achieve chem. accuracy across a broad region of chem. space. It can serve as a valuable resource for the creation of transferable, ready to use potential functions for use in mol. simulations.
    59. 59
      McNutt, A. T.; Bisiriyu, F.; Song, S.; Vyas, A.; Hutchison, G. R.; Koes, D. R. Conformer Generation for Structure-Based Drug Design: How Many and How Good?. J. Chem. Inf. Model. 2023, 63, 6598– 6607,  DOI: 10.1021/acs.jcim.3c01245
      59
      Conformer Generation for Structure-Based Drug Design: How Many and How Good?
      McNutt, Andrew T.; Bisiriyu, Fatimah; Song, Sophia; Vyas, Ananya; Hutchison, Geoffrey R.; Koes, David Ryan
      Journal of Chemical Information and Modeling (2023), 63 (21), 6598-6607CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      Conformer generation, the assignment of realistic 3D coordinates to a small mol., is fundamental to structure-based drug design. Conformational ensembles are required for rigid-body matching algorithms, such as shape-based or pharmacophore approaches, and even methods that treat the ligand flexibly, such as docking, are dependent on the quality of the provided conformations due to not sampling all degrees of freedom (e.g., only sampling torsions). Here, we empirically elucidate some general principles about the size, diversity, and quality of the conformational ensembles needed to get the best performance in common structure-based drug discovery tasks. In many cases, our findings may parallel "common knowledge" well-known to practitioners of the field. Nonetheless, we feel that it is valuable to quantify these conformational effects while reproducing and expanding upon previous studies. Specifically, we investigate the performance of a state-of-the-art generative deep learning approach vs. a more classical geometry-based approach, the effect of energy minimization as a postprocessing step, the effect of ensemble size (max. no. of conformers), and construction (filtering by root-mean-square deviation for diversity) and how these choices influence the ability to recapitulate bioactive conformations and perform pharmacophore screening and mol. docking.
    60. 60
      Foloppe, N.; Chen, I.-J. Energy windows for computed compound conformers: covering artefacts or truly large reorganization energies?. Future Med. Chem. 2019, 11, 97– 118,  DOI: 10.4155/fmc-2018-0400
      60
      Energy windows for computed compound conformers: covering artefacts or truly large reorganization energies?
      Foloppe, Nicolas; Chen, I-Jen
      Future Medicinal Chemistry (2019), 11 (2), 97-118CODEN: FMCUA7; ISSN:1756-8919. (Future Science Ltd.)
      The generation of 3D conformers of small mols. underpins most computational drug discovery. Thus, the conformer quality is crit. and depends on their energetics. A key parameter is the empirical conformational energy window (ΔEw), since only conformers within ΔEw are retained. However, ΔEw values in use appear unrealistically large. We analyze the factors pertaining to the conformer energetics and ΔEw. We argue that more attention must be focused on the problem of collapsed low-energy conformers. That is due to artificial intramol. stabilization and occurs even with continuum solvation. Consequently, the conformational energy of extended bioactive structures is artifactually increased, which inflates ΔEw. Thus, this Perspective highlights the issues arising from low-energy conformers and suggests improvements via empirical or physics-based strategies. Graphical abstr. :.
    61. 61
      Rai, B. K.; Sresht, V.; Yang, Q.; Unwalla, R.; Tu, M.; Mathiowetz, A. M.; Bakken, G. A. Comprehensive Assessment of Torsional Strain in Crystal Structures of Small Molecules and Protein–Ligand Complexes using ab Initio Calculations. J. Chem. Inf. Model. 2019, 59, 4195– 4208,  DOI: 10.1021/acs.jcim.9b00373
      61
      Comprehensive Assessment of Torsional Strain in Crystal Structures of Small Molecules and Protein-Ligand Complexes using ab Initio Calculations
      Rai, Brajesh K.; Sresht, Vishnu; Yang, Qingyi; Unwalla, Ray; Tu, Meihua; Mathiowetz, Alan M.; Bakken, Gregory A.
      Journal of Chemical Information and Modeling (2019), 59 (10), 4195-4208CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The energetics of rotation around single bonds (torsions) is a key determinant of the 3D shape that drug-like mols. adopt in soln., the solid state, and in different biol. environments, which in turn defines their unique phys. and pharmacol. properties. Therefore, accurate characterization of torsion angle preference and energetics is essential for the success of computational drug discovery and design. Here, the authors analyze torsional strain in crystal structures of drug-like mols. in CSD and bioactive ligand conformations in PDB, expressing the total strain energy as a sum of strain energy from constituent rotatable bonds. The authors utilized Cloud computing to generate torsion scan profiles of a very large collection of chem. diverse neutral fragments at DFT(B3LYP)/6-31G*//6-31G** or DFT(B3LYP)/6-31+G*//6-31+G** (for sulfur-contg. mol.). With the data generated from these ab initio calcns., the authors performed rigorous anal. of strain due to deviation of obsd. torsion angles relative to their ideal gas-phase geometries. Contrary to the previous studies based on mol. mechanics, the authors find that in the cryst.-state mols. generally adopt low-strain conformations, with median per-torsion strain energy in CSD and PDB under 1/10th and 1/3rd of a kcal/mol, resp. However, for a small fraction (<5%) of motifs, external effects such as steric hindrance and hydrogen bonds result in strain penalty exceeding 2.5 kcal/mol. The authors find that due to poor quality of PDB structures in general, bioactive structures tend to have higher torsional strain compared to small mol. crystal conformations. However, in the absence of structural fitting artifacts in PDB structures, protein-induced strain in bioactive conformations is quant. similar to those due to the packing forces in small mol. crystal structures. This anal. allows us to establish strain energy thresholds to help to identify biol. relevant conformers in a given ensemble. The work presented here is the most comprehensive study to date that demonstrates the utility and feasibility of gas-phase QM calcns. to study conformational preference and energetics of drug-size mols. Potential applications of this study in computational lead discovery and structure-based design are discussed.
    62. 62
      Taylor, R.; Wood, P. A. A Million Crystal Structures: The Whole Is Greater than the Sum of Its Parts. Chem. Rev. 2019, 119, 9427– 9477,  DOI: 10.1021/acs.chemrev.9b00155
      62
      A Million Crystal Structures: The Whole Is Greater than the Sum of Its Parts
      Taylor, Robin; Wood, Peter A.
      Chemical Reviews (Washington, DC, United States) (2019), 119 (16), 9427-9477CODEN: CHREAY; ISSN:0009-2665. (American Chemical Society)
      A review. The founding in 1965 of what is now called the Cambridge Structural Database (CSD) has reaped dividends in numerous and diverse areas of chem. research. Each of the million or so crystal structures in the database was solved for its own particular reason, but collected together, the structures can be reused to address a multitude of new problems. In this Review, which is focused mainly on the last 10 years, we chronicle the contribution of the CSD to research into mol. geometries, mol. interactions, and mol. assemblies and demonstrate its value in the design of biol. active mols. and the solid forms in which they are delivered. Its potential in other com. relevant areas is described, including gas storage and delivery, thin films, and (opto)electronics. The CSD also aids the soln. of new crystal structures. Because no scientific instrument is without shortcomings, the limitations of CSD research are assessed. We emphasize the importance of maintaining database quality: notwithstanding the arrival of big data and machine learning, it remains perilous to ignore the principle of garbage in, garbage out. Finally, we explain why the CSD must evolve with the world around it to ensure it remains fit for purpose in the years ahead.
    63. 63
      Liebeschuetz, J. W. The Good, the Bad, and the Twisted Revisited: An Analysis of Ligand Geometry in Highly Resolved Protein–Ligand X-ray Structures. J. Med. Chem. 2021, 64, 7533– 7543,  DOI: 10.1021/acs.jmedchem.1c00228
      63
      The Good, the Bad, and the Twisted Revisited: An Analysis of Ligand Geometry in Highly Resolved Protein-Ligand X-ray Structures
      Liebeschuetz, John W.
      Journal of Medicinal Chemistry (2021), 64 (11), 7533-7543CODEN: JMCMAR; ISSN:0022-2623. (American Chemical Society)
      An anal. of the rotatable bond geometry of drug-like ligand models is reported for high-resoln. (<1.1 Å) crystallog. protein-ligand complexes. In cases where the ligand fit to the electron d. is very good, unusual torsional geometry is rare and, most often, though not exclusively, assocd. with strong polar, metal, or covalent ligand-protein interactions. It is rarely assocd. with a torsional strain of greater than 2 kcal mol-1 by calcn. An unusual torsional geometry is more prevalent where the fit to electron d. is not perfect. Multiple low-strain conformer bindings were obsd. in 21% of the set and, it is suggested, may also lie behind many of the 35% of single-occupancy cases, where a poor fit to the e-d. was found. It is concluded that multiple conformer ligand binding is an under-recognized phenomenon in structure-based drug design and that there is a need for more robust crystallog. refinement methods to better handle such cases.
    64. 64
      Tong, J.; Zhao, S. Large-Scale Analysis of Bioactive Ligand Conformational Strain Energy by Ab Initio Calculation. J. Chem. Inf. Model. 2021, 61, 1180– 1192,  DOI: 10.1021/acs.jcim.0c01197
      64
      Large-Scale Analysis of Bioactive Ligand Conformational Strain Energy by Ab Initio Calculation
      Tong, Jiahui; Zhao, Suwen
      Journal of Chemical Information and Modeling (2021), 61 (3), 1180-1192CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      Ligand conformational strain energy (LCSE) plays an important role in virtual screening and lead optimization. While various studies have provided insights into LCSE for small-mol. ligands in the Protein Data Bank (PDB), conclusions are inconsistent mainly due to small datasets, poor quality control of crystal structures, and mol. mechanics (MM) or low-level quantum mechanics (QM) calcns. Here, we built a high-quality dataset (LigBoundConf) of 8145 ligand-bound conformations from PDB crystal structures and calcd. LCSE at the M062X-D3/ma-TZVPP (SMD)//M062X-D3/def2-SVP(SMD) level for each case in the dataset. The mean/median LCSE is 4.6/3.7 kcal/mol for 6672 successfully calcd. cases, which is significantly lower than the ests. based on mol. mechanics in many previous analyses. Esp., when removing ligands with nonarom. ring(s) that are prone to have large LCSEs due to electron d. overfitting, the mean/median LCSE was reduced to 3.3/2.5 kcal/mol. We further reveal that LCSE is correlated with several ligand properties, including formal at. charge, mol. wt., no. of rotatable bonds, and no. of hydrogen-bond donors and acceptors. In addn., our results show that although summation of torsion strains is a good approxn. of LCSE for most cases, for a small fraction (about 6%) of our dataset, it underestimates LCSEs if ligands could form nonlocal intramol. interactions in the unbound state. Taken together, our work provides a comprehensive profile of LCSE for ligands in PDB, which could help ligand conformation generation, ligand docking pose evaluation, and lead optimization.
    65. 65
      Chan, L.; Hutchison, G. R.; Morris, G. M. Understanding Ring Puckering in Small Molecules and Cyclic Peptides. J. Chem. Inf. Model. 2021, 61, 743– 755,  DOI: 10.1021/acs.jcim.0c01144
      65
      Understanding Ring Puckering in Small Molecules and Cyclic Peptides
      Chan, Lucian; Hutchison, Geoffrey R.; Morris, Garrett M.
      Journal of Chemical Information and Modeling (2021), 61 (2), 743-755CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
      The geometry of a mol. plays a significant role in detg. its phys. and chem. properties. Despite its importance, there are relatively few studies on ring puckering and conformations, often focused on small cycloalkanes, 5- and 6-membered carbohydrate rings, and specific macrocycle families. We lack a general understanding of the puckering preferences of medium-sized rings and macrocycles. To address this, we provide an extensive conformational anal. of a diverse set of rings. We used Cremer-Pople puckering coordinates to study the trends of the ring conformation across a set of 140 000 diverse small mols., including small rings, macrocycles, and cyclic peptides. By standardizing using key atoms, we show that the ring conformations can be classified into relatively few conformational clusters, based on their canonical forms. The no. of such canonical clusters increases slowly with ring size. Ring puckering motions, esp. pseudo-rotations, are generally restricted and differ between clusters. More importantly, we propose models to map puckering preferences to torsion space, which allows us to understand the inter-related changes in torsion angles during pseudo-rotation and other puckering motions. Beyond ring puckers, our models also explain the change in substituent orientation upon puckering. We also present a novel knowledge-based sampling method using the puckering preferences and coupled substituent motion to generate ring conformations efficiently. In summary, this work provides an improved understanding of general ring puckering preferences, which will in turn accelerate the identification of low-energy ring conformations for applications from polymeric materials to drug binding.
    66. 66
      Lemm, D.; von Rudorff, G. F.; von Lilienfeld, O. A. Machine learning based energy-free structure predictions of molecules, transition states, and solids. Nat. Commun. 2021, 12, 4468,  DOI: 10.1038/s41467-021-24525-7
      66
      Machine learning based energy-free structure predictions of molecules, transition states, and solids
      Lemm, Dominik; von Rudorff, Guido Falk; von Lilienfeld, O. Anatole
      Nature Communications (2021), 12 (1), 4468CODEN: NCAOBW; ISSN:2041-1723. (Nature Research)
      The computational prediction of atomistic structure is a long-standing problem in physics, chem., materials, and biol. Conventionally, force-fields or ab initio methods det. structure through energy minimization, which is either approx. or computationally demanding. This accuracy/cost trade-off prohibits the generation of synthetic big data sets accounting for chem. space with atomistic detail. Exploiting implicit correlations among relaxed structures in training data sets, our machine learning model Graph-To-Structure (G2S) generalizes across compd. space in order to infer interat. distances for out-of-sample compds., effectively enabling the direct reconstruction of coordinates, and thereby bypassing the conventional energy optimization task. The numerical evidence collected includes 3D coordinate predictions for org. mols., transition states, and cryst. solids. G2S improves systematically with training set size, reaching mean abs. interat. distance prediction errors of less than 0.2 Å for less than eight thousand training structures - on par or better than conventional structure generators. Applicability tests of G2S include successful predictions for systems which typically require manual intervention, improved initial guesses for subsequent conventional ab initio based relaxation, and input generation for subsequent use of structure based quantum machine learning models.
    67. 67
      Hanwell, M. D.; Curtis, D. E.; Lonie, D. C.; Vandermeersch, T.; Zurek, E.; Hutchison, G. R. Avogadro: an advanced semantic chemical editor, visualization, and analysis platform. J. Cheminf. 2012, 4, 17,  DOI: 10.1186/1758-2946-4-17
      67
      Avogadro: an advanced semantic chemical editor, visualization, and analysis platform
      Hanwell, Marcus D.; Curtis, Donald E.; Lonie, David C.; Vandermeersch, Tim; Zurek, Eva; Hutchison, Geoffrey R.
      Journal of Cheminformatics (2012), 4 (), 17CODEN: JCOHB3; ISSN:1758-2946. (Chemistry Central Ltd.)
      Background: The Avogadro project has developed an advanced mol. editor and visualizer designed for cross-platform use in computational chem., mol. modeling, bioinformatics, materials science, and related areas. It offers flexible, high quality rendering, and a powerful plugin architecture. Typical uses include building mol. structures, formatting input files, and analyzing output of a wide variety of computational chem. packages. By using the CML file format as its native document type, Avogadro seeks to enhance the semantic accessibility of chem. data types. Results: The work presented here details the Avogadro library, which is a framework providing a code library and application programming interface (API) with three-dimensional visualization capabilities; and has direct applications to research and education in the fields of chem., physics, materials science, and biol. The Avogadro application provides a rich graphical interface using dynamically loaded plugins through the library itself. The application and library can each be extended by implementing a plugin module in C++ or Python to explore different visualization techniques, build/manipulate mol. structures, and interact with other programs. We describe some example extensions, one which uses a genetic algorithm to find stable crystal structures, and one which interfaces with the PackMol program to create packed, solvated structures for mol. dynamics simulations. The 1.0 release series of Avogadro is the main focus of the results discussed here. Conclusions: Avogadro offers a semantic chem. builder and platform for visualization and anal. For users, it offers an easy-to-use builder, integrated support for downloading from common databases such as PubChem and the Protein Data Bank, extg. chem. data from a wide variety of formats, including computational chem. output, and native, semantic support for the CML file format. For developers, it can be easily extended via a powerful plugin mechanism to support new features in org. chem., inorg. complexes, drug design, materials, biomols., and simulations.
    68. 68
      Avogadro2 Version 1.97. https://two.avogadro.cc/.
      There is no corresponding record for this reference.
    69. 69
      Virtanen, P.; Gommers, R.; Oliphant, T. E.; Haberland, M.; Reddy, T.; Cournapeau, D.; Burovski, E.; Peterson, P.; Weckesser, W.; Bright, J. SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python. Nat. Methods 2020, 17, 261– 272,  DOI: 10.1038/s41592-019-0686-2
      69
      SciPy 1.0: fundamental algorithms for scientific computing in Python
      Virtanen, Pauli; Gommers, Ralf; Oliphant, Travis E.; Haberland, Matt; Reddy, Tyler; Cournapeau, David; Burovski, Evgeni; Peterson, Pearu; Weckesser, Warren; Bright, Jonathan; van der Walt, Stefan J.; Brett, Matthew; Wilson, Joshua; Millman, K. Jarrod; Mayorov, Nikolay; Nelson, Andrew R. J.; Jones, Eric; Kern, Robert; Larson, Eric; Carey, C. J.; Polat, Ilhan; Feng, Yu; Moore, Eric W.; Vander Plas, Jake; Laxalde, Denis; Perktold, Josef; Cimrman, Robert; Henriksen, Ian; Quintero, E. A.; Harris, Charles R.; Archibald, Anne M.; Ribeiro, Antonio H.; Pedregosa, Fabian; van Mulbregt, Paul
      Nature Methods (2020), 17 (3), 261-272CODEN: NMAEA3; ISSN:1548-7091. (Nature Research)
      Abstr.: SciPy is an open-source scientific computing library for the Python programming language. Since its initial release in 2001, SciPy has become a de facto std. for leveraging scientific algorithms in Python, with over 600 unique code contributors, thousands of dependent packages, over 100,000 dependent repositories and millions of downloads per yr. In this work, we provide an overview of the capabilities and development practices of SciPy 1.0 and highlight some recent tech. developments.
    70. 70
      Chan, L.; Hutchison, G. R.; Morris, G. M. Bayesian optimization for conformer generation. J. Cheminf. 2019, 11, 32,  DOI: 10.1186/s13321-019-0354-7
      There is no corresponding record for this reference.
    71. 71
      Chan, L.; Hutchison, G. R.; Morris, G. M. BOKEI: Bayesian optimization using knowledge of correlated torsions and expected improvement for conformer generation. Phys. Chem. Chem. Phys. 2020, 22, 5211– 5219,  DOI: 10.1039/C9CP06688H
      71
      BOKEI: Bayesian optimization using knowledge of correlated torsions and expected improvement for conformer generation
      Chan, Lucian; Hutchison, Geoffrey R.; Morris, Garrett M.
      Physical Chemistry Chemical Physics (2020), 22 (9), 5211-5219CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
      A key challenge in conformer sampling is finding low-energy conformations with a small no. of energy evaluations. We recently demonstrated the Bayesian Optimization Algorithm (BOA) is an effective method for finding the lowest energy conformation of a small mol. Our approach balances between exploitation and exploration, and is more efficient than exhaustive or random search methods. Here, we extend strategies used on proteins and oligopeptides (e.g. Ramachandran plots of secondary structure) and study correlated torsions in small mols. We use bivariate von Mises distributions to capture correlations, and use them to constrain the search space. We validate the performance of our new method, Bayesian Optimization with Knowledge-based Expected Improvement (BOKEI), on a dataset consisting of 533 diverse small mols., using (i) a force field (MMFF94); and (ii) a semi-empirical method (GFN2), as the objective function. We compare the search performance of BOKEI, BOA with Expected Improvement (BOA-EI), and a genetic algorithm (GA), using a fixed no. of energy evaluations. In more than 60% of the cases examd., BOKEI finds lower energy conformations than global optimization with BOA-EI or GA. More importantly, we find correlated torsions in up to 15% of small mols. in larger data sets, up to 8 times more often than previously reported. The BOKEI patterns not only describe steric clashes, but also reflect favorable intramol. interactions such as hydrogen bonds and π-π stacking. Increasing our understanding of the conformational preferences of mols. will help improve our ability to find low energy conformers efficiently, which will have impact in a wide range of computational modeling applications.
    72. 72
      Pordes, R.; Petravick, D.; Kramer, B.; Olson, D.; Livny, M.; Roy, A.; Avery, P.; Blackburn, K.; Wenaus, T.; Würthwein, F. The Open Science Grid. J. Phys. Conf. 2007, 78, 012057,  DOI: 10.1088/1742-6596/78/1/012057
      There is no corresponding record for this reference.
    73. 73
      Sfiligoi, I.; Bradley, D. C.; Holzman, B.; Mhashilkar, P.; Padhi, S.; Wurthwein, F. The Pilot Way to Grid Resources Using glideinWMS. World Congr. Comput. Sci. Inf. Eng. 2009, 2, 428– 432,  DOI: 10.1109/CSIE.2009.950
      There is no corresponding record for this reference.

  • Supporting Information

    Supporting Information

    The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jcim.3c01278.

    • Histograms of the rotatable bonds and conformers in the COD set, comparisons of RMSD and radius of gyration for the COD and Platinum sets, and histograms of torsional angle deviations for GFN2-optimized and ωB97X-D3/def2-SVP-optimized geometries (PDF)


    Terms & Conditions

    Most electronic Supporting Information files are available without a subscription to ACS Web Editions. Such files may be downloaded by article for research use (if there is a public use license linked to the relevant article, that license may permit other uses). Permission may be obtained from ACS for other uses through requests via the RightsLink permission system: http://pubs.acs.org/page/copyright/permissions.html.