Download Hi-Res ImageDownload to MS-PowerPointCite This:J. Chem. Theory Comput. 2019, 15, 2, 1265-1277

Ensemble-Based Replica Exchange Alchemical Free Energy Methods: The Effect of Protein Mutations on Inhibitor Binding

Agastya P. Bhati
Agastya P. Bhati
Centre for Computational Science, Department of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, United Kingdom
More by Agastya P. Bhati
,
Shunzhou Wan
Shunzhou Wan
Centre for Computational Science, Department of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, United Kingdom
More by Shunzhou Wan
http://orcid.org/0000-0001-7192-1999
, and
Peter V. Coveney*
Peter V. Coveney
Centre for Computational Science, Department of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, United Kingdom
*E-mail: [email protected]. Phone: +44 (0)20 7679 4802.
More by Peter V. Coveney
http://orcid.org/0000-0002-8787-7256

Cite this: J. Chem. Theory Comput. 2019, 15, 2, 1265–1277

Publication Date (Web):December 28, 2018

https://doi.org/10.1021/acs.jctc.8b01118

Article Views

2595

Altmetric

Citations

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.

PDF (4 MB)

Get e-Alerts

Supporting Info (2)»Supporting Information Supporting Information

SUBJECTS:

Go to Journal of Chemical Theory and Computation

Get e-Alerts

Abstract

The accurate prediction of the binding affinity changes of drugs caused by protein mutations is a major goal in clinical personalized medicine. We have developed an ensemble-based free energy approach called thermodynamic integration with enhanced sampling (TIES), which yields accurate, precise, and reproducible binding affinities. TIES has been shown to perform well for predictions of free energy differences of congeneric ligands to a wide range of target proteins. We have recently introduced variants of TIES, which incorporate the enhanced sampling technique REST2 (replica exchange with solute tempering) and the free energy estimator MBAR (Bennett acceptance ratio). Here we further extend the TIES methodology to study relative binding affinities caused by protein mutations when bound to a ligand, a variant which we call TIES-PM. We apply TIES-PM to fibroblast growth factor receptor 3 (FGFR3) to investigate binding free energy changes upon protein mutations. The results show that TIES-PM with REST2 successfully captures a large conformational change and generates correct free energy differences caused by a gatekeeper mutation located in the binding pocket. Simulations without REST2 fail to overcome the energy barrier between the conformations, and hence the results are highly sensitive to the initial structures. We also discuss situations where REST2 does not improve the accuracy of predictions.

1. Introduction

ARTICLE SECTIONS

Jump To

Mutations enable proteins to tailor molecular recognition with small-molecule ligands and other macromolecules, and can have a major impact on drug efficacy. Rapid and accurate prediction of drug responses to protein mutations is vital for accomplishing the promise of personalized medicine. The use of targeted therapeutics will benefit cancer patients by matching their genetic profile to the most effective drugs available. Examples of such drugs are gefitinib and erlotinib which belong to a class of targeted cancer drugs called tyrosine kinase inhibitors. A subgroup of patients with nonsmall-cell lung cancer (NSCLC) have specific point mutations and deletions in the kinase domain of epidermal growth factor receptor (EGFR), which are associated with gefitinib and erlotinib sensitivity. Screening for these mutations may identify patients who will have a better response to certain inhibitors.

In silico free energy calculation is one of the most powerful tools to predict the binding affinity of a drug to its target proteins. It employs all-atom molecular dynamics (MD) simulation, a physics-based approach for calculating the thermodynamic properties. The accurate prediction of the binding affinities of ligands to proteins is a major goal in drug discovery and personalized medicine. (1) The use of in silico methods to predict binding affinities has been largely confined to academic research until recently, primarily due to the lack of their reproducibility, as well as lack of accuracy, time to solution, and computational cost.

Recent progress in free energy calculations, marked to some extent by the advent of Schrödinger’s FEP+, (2) has initiated major interest in their potential utility for pharmaceutical drug discovery. The improvements include new sampling protocols in order to accelerate phase space sampling, (3,4) such as Hamiltonian-replica exchange (H-REMD) (5) and its variants, including replica exchange with solute tempering (REST2) (6) and FEP/REST. (7) The replica exchange methods run multiple concurrent (parallel) simulations and occasionally swap information between replicas to improve sampling. For a given set of simulation samples, different free energy estimators (8) can be applied with varying accuracy and precision, of which the multistate Bennett acceptance ratio (MBAR) (9) has become increasingly popular. MBAR makes use of all microscopic states from all of the replica simulations, by reweighting them to the target Hamiltonian. The implementation of an enhanced sampling protocol such as REST2 (6) and the use of the free energy estimator MBAR (9) has been shown to improve the accuracy of the free energy calculations. The rapid growth of computing power and automated workflow tools has also contributed significantly in the wider application of free energy approaches in real world problems.

We have recently developed an approach called thermodynamic integration with enhanced sampling (TIES) (10) which utilizes the concept of ensemble simulation to yield accurate, precise, and reproducible binding affinities. TIES is based on one of the alchemical free energy methods, thermodynamic integration (TI), employing ensemble averages and quantification of statistical uncertainties associated with the results. (11) TIES has already been shown to perform well for a wide range of target proteins and ligands. (10−13) TIES provides a route to reliable predictions of free energy differences meeting the requirements of speed, accuracy, precision, and reliability. The results are in very good agreement with experimental data while the methods are reproducible by construction. Variants of TIES incorporate enhanced sampling techniques REST2 and the free energy estimator MBAR. (11) TIES has been shown to have a positive impact in the drug design process in the pharmaceutical domain. (12,13)

Some protein mutations may fortuitously bring therapeutic benefit to some patients who use a specific drug treatment, while others may impair the ability of a drug to bind with the protein, one of the reasons for the target proteins developing drug resistance. Studying the effect of protein mutations on binding affinity is important for both drug development and for personalized medicine. The purpose of the present paper is to apply the ensemble-based TIES approach (10) to study point mutations in proteins, a variant which we name TIES-PM. TIES-PM employs the TIES methodology to yield rapid, accurate, precise, and reproducible relative binding affinities caused by the protein variants when bound to a ligand.

Here we apply TIES-PM to fibroblast growth factor receptor 3 (FGFR3), one of the four members of the human FGFR family. FGFRs play a critical role in many physiological processes and are recognized therapeutic targets in cancer. (14) Point mutations in FGFRs are among the main genomic alterations, along with fusions and amplifications, contributing to tumor generation and progression. Considerable effort has been dedicated to the development of effective FGFR inhibitors for cancer therapy, some of which are at various phases within clinical trials. (14) In a previous study, (15) we calculated binding free energy differences of inhibitors upon mutations in FGFR1. That study was a critical first step in initiating the TIES protocol. (10) We have also used FGFR1 as one of the molecular systems with which to establish uncertainty quantification within ensemble approaches. (11) In the current study, we consider four FGFR tyrosine kinase inhibitors (TKIs): AZD4547, BGJ-398, JNJ42756493, and TKI258 (see Figure 1), of which the first three are selective and highly potent. (16) Some activating mutations result in distinct changes in drug efficacy. Three single amino acid residue mutations in the kinase domain of FGFR3—V555M, I538V, and N540S—are considered here, of which V555M is the most common gatekeeper mutation. (17) They are among the most frequently observed FGFR3 variants, and confer resistance to these inhibitors in most cases (see the experimental binding affinity changes in Table 1). (16)

Figure 1. Structures of FGFR3 and inhibitors studied in this work: (a) the binding site of tyrosine kinase domain for FGFR3 in complex with ACP, an ATP-analogue (PDB ID: 4K33). FGFR3 is depicted in cartoon and ACP in bond representation. Mutations of three residues, V555, I538, and N540 (ball-and-stick representation), are among the most common genetic variants in FGFR3. The chemical structures of four ATP competitive inhibitors are shown in panels b–e: (b) AZD4547, (c) BGJ-398, (d) TKI258, and (e) JNJ4275649.

Table 1. Relative Binding Free Energies Calculated Using TIES, Incorporating Schemes λR2 and λR2-M as Well as Determined from Experimental K_i Values for All the Inhibitor-Mutant Complexes Studied^a

		ΔΔG_calc
mutant	drug	TIES	λR2^b	λR2-M^b	ΔΔG_exp
V555M	AZD4547-linear	–3.56(0.31)	–2.76(0.12)	–2.70(0.12)	–1.75(0.33)
	AZD4547-bent	0.55(0.41)	–2.07(0.11)	–1.98(0.12)
	BGJ-398	–3.02(0.44)	–3.66(0.12)	–3.60(0.12)	–1.19(0.08)
	TKI258	0.26(0.25)	–1.17(0.13)	–1.11(0.13)	0.97(0.22)
	JNJ42756493	–5.19(0.38)	–3.99(0.16)	–3.92(0.15)	–3.08(0.17)
	MAE	1.75	1.37	1.30
	RMSE	1.84	1.59	1.54
I538V	AZD4547-linear	0.25(0.33)	0.09(0.11)	0.05(0.11)	–2.11(0.32)
	BGJ-398	0.44(0.35)	0.46(0.11)	0.45(0.11)	–0.74(0.21)
	TKI258	–0.65(0.38)	0.47(0.13)	0.38(0.12)	–1.91(0.13)
	JNJ42756493	0.62(0.34)	0.30(0.12)	0.28(0.12)	–2.18(0.10)
	MAE	1.90	2.06	2.02
	RMSE	2.02	2.13	2.08
N540S	AZD4547-linear	–0.43(0.43)	0.91(0.14)	0.95(0.14)	–0.76(0.33)
	BGJ-398	–1.00(0.52)	1.13(0.14)	1.16(0.13)	0.25(0.19)
	TKI258	–1.77(0.60)	1.02(0.14)	1.11(0.14)	–0.90(0.15)
	JNJ42756493	–0.87(0.45)	1.06(0.14)	1.11(0.14)	–1.75(0.21)
	MAE	0.83	1.82	1.87
	RMSE	0.89	1.94	2.00

^{^a}

The mean absolute error (MAE) and root mean square error (RMSE) are also shown for all complexes of each mutant using each free energy scheme. Production runs are 4 ns in all cases. All values are in kcal/mol. The statistical uncertainties associated with each value are shown in brackets.

^{^b}

Highest T_eff for λ-REST2 simulations is 800 K for receptor and 1500 K for complexes in case of mutants I538V and N540S. In the case of mutant V555M, it is 1500 K for the AZD4547 complexes and 600 K for all other complexes; 600 K is used for the receptor.

2. Methods

ARTICLE SECTIONS

Jump To

In this study, ensemble-based λ-REST2 simulations (TIES-λ-REST2) (11) are performed for four TKIs binding to wild-type and mutant FGFR3s (Figure 1). The free energy differences upon mutations are predicted with their associated uncertainties, and compared with experimental data. There are a range of issues and artifacts which affect the reliability and accuracy of MD results. (18) Here we use the latest Amber force fields (see the Simulation Setup section) which are known to be reliable for the present systems, and the same procedures to setup the protein–ligand systems as we recently reported and validated. (10−13)

2.1. Hybrid Topology

A dual topology scheme is used in the current study, similar to that used in our previous studies for the alchemical transmutation of one ligand to another (10) but adapted to handle the transmutation of amino acids in the current study. The reason for this choice is dictated by our use of NAMD. (19) A hybrid residue is introduced, which consists of both the disappearing and appearing amino acids (Figure 2), exclusively belonging to the initial and the final states, respectively. The hybrid potential energy function is set in such a way that the disappearing and appearing parts do not interact with each other. For an alchemical transformation from one amino acid to another, the hybrid structure file is prepared by aligning the mainchain and the common side chain atoms of the appearing residue to those of the disappearing residue.

Figure 2. Different regions in the λ-REST2 simulations. The AZD4547-V555M complex is shown here as an example. The hybrid residue, denoted as the alchemical region, is depicted as a ball-and-stick model. It consists of disappearing (red) and appearing (blue) groups which are slightly separated for reasons of clarity. They can fully or partially overlap in the simulation as there are no interactions between them. The REST2 region, including the alchemical region (red and blue ball-and-stick), part of the ligand (orange bond), and surrounding protein residues (orange stick), is designated as the “hot” region. The selection of the REST2 region is described in the main text (section 2.3 REST2 region).

2.2. Free Energy Schemes

Thermodynamic integration with enhanced sampling (TIES) (10) is used to calculate the free energy differences (ΔΔG_binding) of ligands binding to wild-type and mutant proteins. An alchemical pathway is defined, which corresponds to the transformation of a residue at one end into another at the other end of the pathway. An alchemical coupling parameter, λ, is introduced to define intermediate states with a hybrid potential function V(λ,x), where λ ranges between 0 and 1 corresponding to the initial and final states, respectively. In thermodynamic integration, the alchemical free energy change ΔG_alch is given by the following equation:

(1)

where ⟨∂V(λ,x)/∂λ⟩_λ denotes an ensemble average of the potential energy derivative in state λ. Ensemble MD simulations are run at each λ window for both apo protein and ligand–protein complex. Equation 1 is evaluated using a stochastic integration method since the integrand is generated from a Gaussian random process, (20) and the associated uncertainty is calculated accordingly. (10) The free energy difference ΔΔG_binding is then calculated as the difference of the alchemical free energy changes ΔG_alch of apoprotein and complex, and the uncertainty as the propagation of the errors. Three schemes (11) are used in the current study, as (i) the standard TIES, (10) (ii) an ensemble of λ-REST2 simulations termed as TIES-λ-REST2 (λR2), and (iii) TIES-λ-REST2 with MBAR estimator termed TIES-λ-REST2-M (λR2-M). (11) All of the three schemes use ensemble-based simulations. (10,11) In standard TIES, simulations are run indepedently at each predefined λ value and at a constant temperature (300 K). In TIES-λ-REST2 simulations, a predefined number of parallel REST2 replicas are run with regular exchange attempted between neighboring replicas of which both the alchemical parameters λ and the effective temperatures T_eff differ. The calculated binding free energy differences from these schemes are denoted as ΔΔG_calc^TIES, ΔΔG_calc^λR2 and ΔΔG_calc^λR2–M, respectively, which all are obtained from eq 1 but differ in the ways of deriving the integrands. In standard TIES (i), the potential energy in the integrand is a function of λ (the temperature is a constant), and the average includes samples from a specific λ window. In λ-REST2 (ii), the potential energy is a function of (λ, T_eff), and the average includes samples from a specific λ window. In λ-REST2-M (iii), the potential energy is a function of (λ, T_eff), and the average includes samples from multiple λ windows using MBAR.

2.3. REST2 Region

As described in Bhati et al., (10) a small region of the molecular system is designated as the so-called “hot” region for all λ-REST2 simulations (Figure 2). This region is usually referred to as the REST2 region. It is critical to properly define the region for the REST2 simulations in order to improve the sampling of conformations relevant to binding. If the region is too small, the overall impact of applying REST2 may be insufficient to prevent the molecule from getting trapped in one or more local energy minima. It has been shown (21) that using the default FEP+ protocol, (2) in which only perturbed ligand groups are included in the REST2 region, is not sufficient for some cases to obtain proper sampling. Another study (22) shows that choosing only a mutant residue as the hot region has no effect on binding free energy prediction in most cases. On the other hand, when the region is too big, a large number of replicas within the replica exchange process may be required to cover a given range in effective temperatures, while the sampled conformations may not be relevant to stable binding of the inhibitor at all. It has been suggested to include key protein residues within the REST2 regions, which are identified either by visual inspection (21) or by analyses of preliminary simulations performed prior to FEP+ runs. (23)

In this study, the REST2 region for all mutations is defined as follows: for unbound protein calculations, it includes the mutant residue and all protein residues within 3 Å distance of the former; for bound protein calculations, it comprises the mutant residue, all protein residues within 3 Å of the mutant residue and 4 Å of the ligand, and all ligand atoms within 4 Å of the mutant residue. The nonbonded interactions of the atoms in the “hot” region are reduced by a scaling factor based on an effective temperature (T_eff). The alchemical region (Figure 2), which is part of the “hot” region, is also scaled by the alchemical coupling parameter, λ. The λ value increases linearly from 0 to 1 with replicas, whereas T_eff varies such that it attains its maximum at the midpoint and decreases to 300 K at the end-points. Samples from a specific REST2 replica are used to calculate ∂V/∂λ at that state for each λ-REST2 simulation followed by standard TIES analysis to yield ΔG_alch and its associated uncertainty. (10)

2.4. Simulation Setup

The structure of FGFR3 was taken from the protein data bank (PDB ID: 4K33, Figure 1). The missing residues in the file were built by ModLoop, (24) and the mutant/engineered residues were restored to their wild-type forms. The inhibitors were manually positioned into the binding site on the basis of their existing X-ray structures as follows: BGJ-398 from PDB ID 3TT0, JNJ42756493 from PDB ID 5EW8, TKI258 from PDB ID 5AM7, and AZD4547 from PDB IDs 4V05 and 4RWK as there are two distinct conformations for it, denoted as “linear” and “bent” in the current study. (25) The crystal water molecules of 4K33 were retained unless they overlapped with the aligned inhibitors. The inhibitors were optimized at the Hartree–Fock level with the 6-31G* basis (HF/6-31G*) in Gaussian 03 (26) and parametrized using Antechamber and restrained electrostatic potential (RESP) modules in AmberTools 17 (27) with the general AMBER force field (GAFF). (28) The Amber ff14SB force field (29) was used for the protein. All systems were solvated in orthorhombic water boxes with a minimum extension from the protein of 14 Å. The TIP3P water model was used. The molecular systems were neutralized with Na⁺ or Cl^– ions.

2.5. Simulations

The customized version of the NAMD 2.11 package, (19) with implementation of REST2 for alchemical simulations, (30) was used for all the TIES-PM simulations. The systems were maintained at a temperature of 300 K and a pressure of 1 bar in an NPT ensemble. A time step of 2 fs was used. We used the protocol established in our previous publications (10,11) in which an ensemble of five replicas was used; 2 ns of equilibration and 4 ns of production were conducted for each replica. To check the convergence of the calculated free energies, some simulations were extended up to 20 ns. A soft core potential was used for the van der Waals interactions which were scaled up/down linearly across the full λ range (0 to 1) for the appearing/disappearing atoms, respectively. The electrostatic interactions of the disappearing atoms were linearly decoupled from the simulations between λ values of 0 and 0.55 and completely turned off beyond that, while those of the appearing atoms were not turned on until λ = 0.45, and then linearly coupled to the simulations from λ value 0.45 to 1. An exchange of configurations between two neighboring λ replicas was attempted every 1 ps, and conformations were saved every 10 ps.

2.6. Computational Resources

The TIES-λ-REST2 simulations require a large number of MD runs to be performed. On modern large scale high performance computers, all simulations can be run in parallel and completed in the same wall clock time as needed for a single MD simulation. All simulations were run on the BlueWaters supercomputer at the National Center for Supercomputing Applications of the University of Illinois at Urbana–Champaign (https://bluewaters.ncsa.illinois.edu) and Titan at Oak Ridge National Laboratory (https://www.olcf.ornl.gov/olcf-resources/compute-systems/titan/). For a 2 ns equilibration and 4 ns production MD simulation, it took 14.6 h on 4 nodes (128 cores) of BlueWaters, and 8.7 h on 15 nodes (240 cores) of Titan. Collectively about 27.8 million core hours were consumed in the course of this study.

3. Results

ARTICLE SECTIONS

Jump To

Table 1 contains the calculated as well as experimental relative binding affinities (ΔΔG) for all the mutant-inhibitor complexes studied. ΔΔG_calc values from three free energy schemes, namely TIES, (10) TIES-λ-REST2, and TIES-λ-REST2-M, (11) are reported. Some significant improvements are observed from TIES-λ-REST2 simulations, while the inclusion of MBAR only slightly improves the accuracy and precision (Table 1). In the following analyses, we mainly focus on the comparisons of TIES and TIES-λ-REST2. The experimental values are determined using the K_i values reported by Patani et al. (16) Mean absolute error (MAE) and root-mean-square error (RMSE) values for all complexes of every mutant using each free energy scheme are included as a measure of the accuracy of the simulation results. The inhibitor AZD4547 has been reported to bind with the FGFR kinase gatekeeper mutant in two distinct conformations—linear and bent—experimentally. (25) Results for both are shown in Table 1. It should be noted that the FGFR3 gatekeeper mutation V555M occurs inside the binding pocket (“local”), while the other two mutations (I538V and N540S) occur away from the binding pocket (“remote”). The effect of local mutations on the binding of inhibitors can be attributed to the changes in the local environment of the binding pocket altering the direct interaction between protein and inhibitor. On the other hand, remote mutations do not have any direct interaction with the inhibitor and hence can be expected to affect the inhibitor binding through indirect means such as large scale conformational changes in the protein. Such events may occur on a time scale of the order of μs to ms. Below, we discuss the results from these two categories of mutations separately.

3.1. Local Mutation

In the case of the V555M mutant, ΔΔG_calc^TIES predicts the resistance behavior of all inhibitors correctly except AZD4547 starting from the bent conformation; that is, the predicted ΔΔG_calc values have the same signs as those of the corresponding experimental values ΔΔG_exp (Figure 3). In other words, the predictions agree directionally with the experimental values. We find that, for standard TIES, the accuracy of the predictions is not very good, most of the complexes having an absolute error close to 2 kcal/mol with a MAE and RMSE of 1.75 and 1.84 kcal/mol, respectively. In addition, the predicted relative binding affinity of the AZD4547–V555M complex is sensitive to its initial structure. The ΔΔG_calc values for the linear and bent conformations of AZD4547 differ by about 4 kcal/mol. It has been shown experimentally that AZD4547 coexists in its linear as well as bent conformations only when binding to the gatekeeper mutant. (25) This means that, during an alchemical transformation from valine to methionine (V555M), AZD4547 should exhibit only its linear conformation at the λ = 0 end-point (V555), but have an increasingly mixed population of both linear and bent conformations on approaching the λ = 1 end-point (M555). It appears that the energy barrier between these two conformations is too high to be overcome using standard MD simulations at 300 K. Thus, the inhibitor remains trapped in its starting conformation throughout a standard TIES calculation. This explains why the TIES prediction ΔΔG_calc^TIES is so sensitive to its initial structure and does not agree directionally with its experimental value in the case of the bent conformation of AZD4547.

Figure 3. Comparison of the predicted ΔΔG_calc values using TIES (black circles), TIES-λ-REST2 (λR2, up/down triangles) with those from experiments for V555M mutant complexes with the highest T_eff of the chosen REST2 region at 600 K for receptor and complexes except those with AZD4547 which are at 1500 K (red triangles pointing up), and at 1500 K for receptor and 3000 K for complexes (blue triangles pointing down). Results of AZD4547 from the bent conformation are represented using filled circles and triangles. The dotted lines (x = 0 and y = 0) create four quadrants. Data points in quadrants I (x > 0 and y > 0) and III (x < 0 and y < 0) indicate that the calculated binding free energy differences agree directionally with those from the experimental data. The results from TIES-λ-REST2-M (λR2-M) are very close to those from λR2 (Table 1), and are not shown for reasons of clarity.

To overcome the large energy barrier between the two conformations of AZD4547 and also to study the effect of the accelerated sampling method, λ-REST2, (7) on ΔΔG predictions in general, we performed TIES-λ-REST2 simulations to get ΔΔG_calc^λR2 and ΔΔG_calc^λR2–M (refer to Table 1 and Figure 3). On comparing them with ΔΔG_calc^TIES, we find an overall improvement in the relative binding affinity predictions, the MAE and RMSE reducing by 0.38 and 0.25 kcal/mol with scheme λR2 and by 0.45 and 0.30 kcal/mol with scheme λR2-M, respectively. The AZD4547, both for the linear and the bent conformations, benefit from REST2 with their ΔΔG_calc predictions improving drastically. However, it is clear from Figure 3 that, out of the five complexes (including the two conformations of AZD4547), ΔΔG predictions for only three complexes improve using TIES-λ-REST2. The relative binding affinity for the V555M-BGJ-398 complex remains the same as in the case of standard TIES while that for the V555M–TKI258 complex is less accurate than standard TIES using TIES-λ-REST2. This is because some conformations are sampled in the TIES-λ-REST2 simulations that are irrelevant to stable ligand binding and lead to the deviations of the predictions from the experimental values (see more details in the Discussion section).

To investigate the effects of the highest REST2 temperature (T_eff) on the predictions, we increased T_eff value from 600 to 1500 K for the receptor and 600/1500 to 3000 K for the complexes. As can be clearly seen from Figure 3, increasing the temperature of the “hot” region improves the accuracy of the results in most cases and reduces MAE and RMSE by up to 0.6 kcal/mol (Table 2). Three out of the five inhibitors bound to the V555M mutant then generate predictions closer to experiment by more than 0.7 kcal/mol (BGJ-398, by 1.74 kcal/mol; JNJ42756493, by 0.71 kcal/mol; AZD4547-linear, by 0.91 kcal/mol). Although the absolute error (0.68 kcal/mol) increases slightly for the AZD4547-bent using the higher T_eff, it is still well on the level of accuracy expected from such alchemical approaches. (11) Increasing the temperature allows greater flexibility within the system being simulated and facilitates access to key regions of phase space by allowing high energy barriers in the potential energy surface to be crossed. The inhibitor TKI258, when bound to the V555M mutant, remains an exception as its predicted relative free energy is displaced from the perfect correlation line on increasing the temperature. This exceptional behavior of TKI258 is due to an even higher population of irrelevant conformations sampled in the simulations when a higher temperature is used (see the Discussion section).

Table 2. Relative Binding Free Energies Calculated Using Schemes λR2 and λR2-M with Different Highest Effective Temperature (T_eff) Compared with Corresponding Experimental Values for All the Inhibitor-Mutant Complexes Studied^a

		ΔΔG_calc
mutant	drug	λR2^b	λR2-M^b	λR2^c	λR2-M^c	ΔΔG_exp
V555M	AZD4547-linear	–2.76(0.12)	–2.70(0.12)	–1.85(0.07)	–1.82(0.06)	–1.75(0.33)
	AZD4547-bent	–2.07(0.11)	–1.98(0.12)	–1.07(0.07)	–1.11(0.06)
	BGJ-398	–3.66(0.12)	–3.60(0.12)	–1.92(0.08)	–1.96(0.06)	–1.19(0.08)
	TKI258	–1.17(0.13)	–1.11(0.13)	–1.41(0.08)	–1.42(0.06)	0.97(0.22)
	JNJ42756493	–3.99(0.16)	–3.92(0.15)	–2.88(0.12)	–2.87(0.11)	–3.08(0.17)
	MAE	1.37	1.30	0.82	0.82
	RMSE	1.59	1.54	1.16	1.16
I538V	AZD4547-linear	0.09(0.11)	0.05(0.11)	–0.12(0.08)	–0.04(0.07)	–2.11(0.32)
	BGJ-398	0.46(0.11)	0.45(0.11)	0.01(0.08)	0.09(0.07)	–0.74(0.21)
	TKI258	0.47(0.13)	0.38(0.12)	0.01(0.08)	0.12(0.08)	–1.91(0.13)
	JNJ42756493	0.30(0.12)	0.28(0.12)	–0.01(0.07)	0.11(0.07)	–2.18(0.10)
	MAE	2.06	2.02	1.71	1.80
	RMSE	2.13	2.08	1.80	1.89
N540S	AZD4547-linear	0.91(0.14)	0.95(0.14)	0.72(0.11)	0.74(0.11)	–0.76(0.33)
	BGJ-398	1.13(0.14)	1.16(0.13)	0.67(0.11)	0.67(0.11)	0.25(0.19)
	TKI258	1.02(0.14)	1.11(0.14)	0.71(0.12)	0.72(0.12)	–0.9(0.15)
	JNJ42756493	1.06(0.14)	1.11(0.14)	0.72(0.12)	0.72(0.12)	–1.75(0.21)
	MAE	1.82	1.87	1.50	1.50
	RMSE	1.94	2.00	1.66	1.67

^{^a}

The mean absolute error (MAE) and root mean square error (RMSE) for all complexes of each mutant using each free energy scheme are also shown. Production runs are 4 ns in all cases. All values are in kcal/mol. Statistical uncertainties associated with each value are shown in the brackets.

^{^b}

^{^c}

Highest T_eff for λ-REST2 simulations is 1500 K for receptor and 3000 K for complexes.

3.2. Remote Mutations

In the case of remote mutations, the calculated relative free energies do not agree well with the experimental values (Table 1 and Figure 4). This is not surprising given that the predictions are made using 4 ns long simulations. As mentioned earlier, the effect of remote mutants on the binding of an inhibitor is not due to a change in the direct interaction between the inhibitor and the protein. It generally involves a change in the protein conformation which indirectly affects the binding of the inhibitor. Such conformational changes are close to impossible to capture with molecular dynamics simulations of short temporal duration. This is further confirmed by the fact that the predicted ΔΔG_calc^TIES values for all inhibitors (except TKI258 whose unusual behavior is further discussed in the next section) using standard TIES are very close to each other in the cases of both the I538V as well as the N540S mutant. This essentially means that short duration simulations are only able to capture the changes in the immediate vicinity of the alchemical transformation (i.e., the mutation in this case).

Figure 4. Comparison of the predicted ΔΔG_calc values using TIES (black circles), TIES-λ-REST2 (λR2, up/down triangles) with those from experiment for for all inhibitors bound to FGFR3: (a) I538V mutant and (b) N540S mutant, when the highest T_eff for the chosen REST2 region is at 800 K for receptor and 1500 K for complexes (red triangles pointing up) and at 1500 K for receptor and 3000 K for complexes (blue triangles pointing down). The dotted lines (x = 0 and y = 0) create four quadrants. Data points in quadrants I (x > 0 and y > 0) and III (x < 0 and y < 0) indicate that the ΔΔG_calc values agree directionally with ΔΔG_exp. The results from TIES-λ-REST2-M (λR2-M) are very close to those from λR2 (Table 1), and are not shown for reasons of clarity.

The predicted ΔΔG_calc values for complexes with the I538V mutant using all three free energy schemes are statistically close to zero. Among the three mutations investigated, I538V is the most distant from the bound inhibitors. The I538V mutation involves alchemically transforming isoleucine to valine which are both nonpolar amino acids. Thus, this transformation does not significantly affect the charge distribution of the protein and hence also does not affect its long-range interactions. The mutation does not directly affect the two calculated ΔG_alch values (eq 1) in the presence and absence of an inhibitor, from which the free energy difference ΔΔG_calc is calculated. The experimentally detected ΔΔG_exp must be generated from a mechanism which is not captured in the simulations.

Although the ΔΔG_calc values are not close to zero for the N540S mutation, the predictions are consistently positive. This means that the two calculated ΔG_alch values for the alchemically transforming protein residue differ in the presence and absence of an inhibitor. In the case of remote mutations, the two ΔG_alch values can differ when there is a considerable change in long-range interactions of the mutant with the inhibitor. The N540S mutation involves changing asparagine to serine (effectively −CONH₂ to −OH) which alters the charge distribution of the protein as well as its long-range interactions. Another reason for the nonzero ΔΔG_calc predictions in this case is that the mutation is closer to the inhibitors as compared to the I538V mutant. The shortest distance between the hybrid N540S residue and the inhibitor is 6 Å while for the I538V mutation it is 9 Å. Thus, one would expect that there would be a greater likelihood of standard TIES being able to capture the effect of the N540S than the I538V mutant. Indeed, it should be noted that the complexes bound to the N540S mutant in Table 1 have the least MAE/RMSE among all mutants with standard TIES predictions.

We also performed TIES-λ-REST2 simulations of duration up to 4 ns for the remote mutations but they do not improve the accuracy of results. This may be because the indirect mechanisms which potentially affect the inhibitor binding in such cases occur on time scales longer than can be computed by the simulations performed in this study and hence the “correct” region of the phase space is not sampled even using TIES-λ-REST2 simulations. Remote mutations may modulate the inhibitor–protein interactions via induced allosteric conformational changes occurring over a wide range of space and time scales. A number of computational methodologies have been developed for modeling large-scale motions of proteins, including coarse-grained molecular dynamics and accelerated MD. The prediction of binding free energies may be improved by taking into account all of the conformations, with statistical reweighting techniques to optimally merge data obtained from the enhanced approaches. It is interesting to note in Figure 4 that the ΔΔG_calc^{λR2/λR2–M} values for all inhibitors are close to each other for both the remote mutations unlike in the case of standard TIES, for which TKI258 was an exception. Thus, TIES-λ-REST2 brings all ΔΔG_calc values to the same baseline irrespective of the inhibitor bound.

Increasing the temperature of the “hot” region, from 800 to 1500 K for the receptor and 1500 to 3000 K for the complexes, improves the accuracy of the results and reduces MAE and RMSE of predicted ΔΔG_calc for both of the remote mutants (Table 2). However, the predictions do not agree with the experimental values (Figure 4). In contrast with the observation for the local mutation (section 3.1), the ΔΔG calculations do not benefit from REST2 for the remote mutants studied here.

3.3. Effect of Extending Simulation Time

As we explained earlier, short duration simulations are usually unable to correctly predict the relative binding free energies of remote mutations. In this section, we present the outcome of our attempts to extend the duration of simulations. Figure 5 displays the variation of cumulative average of ΔΔG_calc values using standard TIES with the length of simulation extended up to 20 ns for the complexes involving I538V mutant. Apart from small variations, the predicted ΔΔG_calc values remain constant and do not exhibit any signs of getting closer to the corresponding experimental values for all inhibitors except JNJ42756493. In the case of JNJ42756493, the ΔΔG_calc value seems to be drifting toward the experimental value but is still quite far from it after 20 ns. This suggests that 20 ns of standard MD simulation is not sufficiently long to sample the relevant conformations of the complexes involving remote mutants.

Figure 5. Variation of the cumulative average of ΔΔG_calc^TIES with simulation length for all inhibitors bound to the FGFR3 I538V mutant. The corresponding experimental value for each inhibitor is shown by a dashed line of the same color.

We also extended the TIES-λ-REST2 simulations to see if “heating” a local portion of the complexes around the mutant residue and/or the binding pocket has any impact on the predicted free energies. Figure 6 displays the variation of cumulative average of the calculated ΔΔG_calc values with the simulation length up to 20 ns for all complexes of the three mutants. In the case of complexes involving the V555M mutant, the general trend is that the predicted ΔΔG_calc values get closer to the experimental values. However, it should be noted that, out of the five complexes, the largest difference between the predicted values of ΔΔG_calc at 4 and 20 ns is 0.75 kcal/mol for the V555M–JNJ42756493 complex (see the black line in Figure 6; from −4.18 kcal/mol at 4 ns to −3.43 kcal/mol at 20 ns). The corresponding values for the other V555M complexes are less than or equal to 0.5 kcal/mol. This is a marginal gain measured against the expense of the computation leading to a very high cost–benefit ratio. This observation is in line with our previous studies where we have shown that “long” simulations furnish little to no advantage when the alchemical transformation is local, that is, when it occurs in the binding site. (10,11)

Figure 6. Variation of the cumulative average of ΔΔG calculated using schemes λR2 and λR2-M with simulation length for all complexes. The highest T_eff for receptor and complex are 1500 and 3000 K, respectively, in the case of I538V and N540S mutants, while the corresponding values in the case of V555M mutant are 600 and 600 K/1500 K. The corresponding experimental values for each inhibitor are shown by a dashed line of the same color.

Unlike the V555M mutant, in the case of remote mutants, we extended the λ-REST2 simulations with the highest T_eff of 1500 K for receptor and 3000 K for complexes up to 20 ns. The length of the simulation does not affect the predicted ΔΔG_calc values in these cases either. The predictions remain quite stable, consistently away from the experimental values and close to each other for all complexes involving remote mutations.

4. Discussion

ARTICLE SECTIONS

Jump To

In this section, we discuss how the application of λ-REST2 may be useful in some cases while degrading the quality of results in others. We provide details on the variation in the predicted ΔΔG values for the V555M–AZD4547 complex with the two conformations of the inhibitor AZD4547 using the standard TIES and then how λ-REST2 simulations bring them closer. We also explain the anomalous behavior of the inhibitor TKI258 in detail and provide evidence for how “heating” adversely affects the results in this case. On the basis of the discussion in this section, we formulate some caveats and recommendations concerning the application of the λ-REST2 technique in general for free energy predictions.

4.1. Improved Sampling of AZD4547 on “Heating”

As noted earlier, the inhibitor AZD4547 has been found to bind with the FGFR gatekeeper mutation in two distinct conformations—linear and bent—as shown in Figure 7. However, in the case of the wildtype (WT) FGFR kinase and its other mutants, AZD4547 occurs only in the linear conformation. This suggests that while simulating an alchemical transformation corresponding to the FGFR3 WT to V555M mutant, the inhibitor AZD4547 should occur only in the linear conformation at λ = 0 end-point (WT), while adopting an increasingly mixed population of both conformations on approaching λ = 1 end-point (V555M). In this section, we quantify the occurrence of both conformations of AZD4547 among the MD trajectories of the V555M–AZD4547 complex in the case of different free energy schemes and discuss its impact on the predicted relative free energies.

Figure 7. Two distinct conformations of inhibitor AZD4547 found experimentally when bound to the FGFR gatekeeper mutant. The three hydrogen bonds, marked with black dashed lines and labeled as H1, H2, and H3, keep the middle portion of the inhibitor stable. The value of the dihedral angle between the four carbon atoms highlighted in orange can be used as an indicator of the occurrence of the two conformations. The atoms displayed as balls lie in the REST2 region while the ones displayed as lines reside outside it.

In Figure 7, the three hydrogen bonds, which AZD4547 forms with the glutamic acid and alanine residues of the hinge region of FGFR kinase, are marked with black dashed lines and labeled as H1, H2, and H3. These hydrogen bonds keep the middle portion of AZD4547 stable. Figure 8 displays the normalized frequency distributions of these three hydrogen bonds in the λ-REST2 trajectories at the λ = 1 end-point of V555M–AZD4547 complexes with the linear as well as the bent starting structures. H1 and H2 peak around 2 Å while H3 peaks around 2.5 Å. This makes it clear that AZD4547 remains stably bonded to the hinge region of the FGFR3 V555M mutant. The four carbon atoms which connect this stable middle portion of AZD4547 with its free head portion are highlighted in orange in Figure 7. It can be seen that the dihedral angle between these four carbon atoms may be used as a reliable indicator of the type of conformation AZD4547 exists in at a given point in the MD trajectory. Therefore, we use this information to quantify the occurrence of the two conformations of AZD4547.

Figure 8. Normalized frequency distributions of the three hydrogen bonds (H1, H2, and H3 from Figure 7) in λ-REST2 trajectories at the λ = 1 end-point of the V555M–AZD4547 complexes when using the linear as well as the bent starting structures.

Figure 9 displays the normalized frequency distributions of the dihedral between the four carbon atoms highlighted in orange in Figure 7 from the standard TIES as well as λ-REST2 trajectories at λ = 0, 0.5, and 1 for V555M–AZD4547 complexes with both the linear and the bent starting structures. The peaks centered around +160° correspond to the linear conformation whereas the peaks around −80° correspond to the bent conformation. It is easy to recognize from Figure 9 that, in the case of standard TIES (shown in blue), the type of conformations sampled is entirely dependent on the starting structure of the complex and that there is absolutely no mixing of the states during such simulations. Thus, there are negligible peaks around −80° and +160° when starting with the linear and the bent conformations, respectively. This explains why the predicted ΔΔG values are sensitive to the starting structure and are very different for the two different starting structures using TIES. The plots in black and red denote the distributions from the first and the last 4 ns of the 20 ns long λ-REST2 trajectories. It is evident that λ-REST2 allows sampling the states from both conformations irrespective of the starting structure. This is possible through regular exchanges of conformations between the neighboring states and heating of the REST2 region in the intermediate states. As becomes obvious from Figure 9, the distributions at λ = 0.5 are relatively smoother with non-negligible proportions of both the conformations. However, the picture is not so simple in the case of end-points as discussed next.

Figure 9. Normalized frequency distributions of the dihedral angle between the four carbon atoms highlighted in orange in Figure 7 for different λ states of V555M–AZD4547 complexes in standard TIES (in blue) as well as λ-REST2 simulations showing the relative populations of the two conformations of AZD4547. In the case of λ-REST2 simulations, the distributions from the first (1–4 ns; in black) and the last 4 ns (17–20 ns; in red) are shown separately.

Ideally, the λ = 0 end-point should have samples predominantly if not exclusively from the linear conformations while the λ = 1 end-point should have samples from both the conformations. However, we find that there is some mixing during the first 4 ns of λ-REST2 simulations which is not enough to reach the ideal situation and hence a dependence on the starting structure persists. In the case of the linear starting structure, there are predominantly linear conformations even at the λ = 1 end-point during the first 4 ns. Similarly, in the case of the bent starting structure, predominantly the bent conformations persist even at the λ = 0 end-point. However, during the last 4 ns of λ-REST2 simulations, there is a noticeable improvement in both situations. In the case of the linear starting structure, there is a visible peak around −80° at λ = 1 end-point. In the case of the bent starting structure, there is an almost equal proportion of both conformations at λ = 1 end-point. Moreover, the peak around −80° is much smaller as compared to the first 4 ns at λ = 0 end-point. It is interesting to note that the −80° peak is always lower for the λ = 0 end-point as compared to the λ = 1 end-point in the case of λ-REST2 simulations. Indeed, the switching from bent to linear conformation appears to be easier than the converse through λ-REST2 simulations. This is because both of the end-points accept the linear conformation, while only the λ = 1 end-point tolerates the bent conformation. Through this selective pressure, the linear conformation is more likely to spread than the bent one during the replica exchange simulations.

4.2. The Exceptional Case of TKI258: Limitations of λ-REST2

The inhibitor TKI258 is an anomaly in this study. As is clear from Figure 4 and Table 1, its ΔΔG prediction consistently becomes less accurate on applying λ-REST2 as compared to the standard TIES in case of all three mutants. The absolute errors of the TKI258 complexes increase by 1.43, 1.12, and 1.05 kcal/mol when bound with V555M, I538V, and N540S, respectively, on using λ-REST2. In addition, its relative binding affinity predictions using λR2 as well as λR2-M become less accurate on increasing the highest T_eff to 3000 K in the case of the V555M mutant. In this section, we provide an explanation for such behavior of TKI258. Figure 10 displays the binding pose of TKI258 found experimentally. It forms two hydrogen bonds with glutamic acid and alanine residues of the hinge region of the protein (labeled as H1 and H2 in Figure 10). Figures 11, 12, and 13 show the normalized frequency distributions of H1 and H2 for λ = 0 and λ = 1 end-points of TKI258 complexes in the case of the standard TIES as well as λ-REST2 simulations. Below we discuss them in detail.

Figure 10. Inhibitor TKI258 bound to V555M mutant. It forms two hydrogen bonds with the hinge region of the protein which are displayed with black dashed lines and labeled as H1 and H2. The atoms shown as balls lie in the “hot” region. The atoms are shown in the standard color code: carbon in green, oxygen in red, nitrogen in blue, hydrogen in white, and fluorine in pink.

Figure 11. Normalized frequency distributions of H1 and H2 from Figure 10 for the two end-points in the case of the standard TIES as well as λ-REST2 simulations of the V555M–TKI258 complex. λ-REST2 simulations sample a larger comformational space than TIES, as evidenced by the lower and wider distributions of the distances, and the second peaks in the λ = 1 end-point (V555M).

Figure 12. Normalized frequency distributions of H1 and H2 from Figure 10 for the two end-points in the case of standard TIES as well as λ-REST2 simulations of the I538V–TKI258 complex. The long tails and additional peaks beyond 4 Å indicate that λ-REST2 simulations sample some conformations irrelevant to stable inhibitor binding.

Figure 13. Normalized frequency distributions of H1 and H2 from Figure 10 for the two end-points in the case of standard TIES as well as λ-REST2 simulations of the N540S–TKI258 complex. The long tails and additional peaks beyond 4 Å indicate that λ-REST2 simulations sample some conformations irrelevant to stable inhibitor binding.

Figure 11 compares the normalized frequency distributions of H1 and H2 in the ensemble of conformations for the two end-points (λ = 0 and λ = 1) of the standard TIES calculation as well as λ-REST2 calculation. Both H1 and H2 have peaks around 2 Å with almost negligible frequencies between 3 and 4 Å and vanish for distances greater than 4 Å. This is true for both the end-points which suggests that the inhibitor remains quite stable and tightly bound to the protein via the two hydrogen bonds throughout the simulations at both the end-points. This explains why ΔΔG_calc^TIES for the V555M–TKI258 complex is close to zero.

On the other hand, in the case of λ-REST2 simulations, the right-hand tails of the H1 and H2 distributions extend to 7 Å. However, in the case of the λ = 0 end-point, they have negligible frequencies beyond 3 Å unlike the λ = 1 end-point where there are small peaks for both H1 and H2 around 4 and 5 Å, respectively. This happens because, due to the “heating”, the inhibitor drifts away from the protein and hence binds only weakly to it at one of the end-points as compared to the other end-point. This explains the negative value of ΔΔG_calc^λR2(−M) for the V555M–TKI258 complex. The important thing to note here is that the “heating” in the intermediate λ-windows and regular exchanges between the neighboring REST2-replicas causes the complex to visit some “unwanted” high energy conformations leading to degraded results. The smaller peaks of the H1 and H2 distributions at λ = 1 end-point suggest that the V555M–TKI258 complex has a higher energy minimum which is not sampled by standard MD simulations and is probably not observed experimentally to the extent realized by the λ-REST2 simulations (X-ray structures show that all reversible ATP-competitive inhibitors form one or more stable H-bond(s) with the hinge region of the receptors). The more heating there is, the greater is the population of this higher energy minimum and hence the more negative is the ΔΔG_calc prediction (refer to Figures 3, 4, and Table 2).

Comparing Figure 11 with Figure 8, it is clear that such a drifting of the inhibitor does not occur in the case of V555M–AZD4547 complex even though the highest T_eff for the latter is 1500 K against 600 K in the case of the former. To find out if a similar process arises in other complexes, we performed a hydrogen-bond analysis for all complexes whereby we determined the occupancies of all the hydrogen bonds formed between glutamic acid and alanine residues of the hinge region of the protein and the inhibitor (refer to Table S7 of the Supporting Information; the bond and angle cut-offs used to determine the occurrence of the hydrogen bonds were 3 Å and 135°). We found that, except for JNJ42756493, all inhibitors form one or more hydrogen bonds with at least one of the two residues such that their occupancies are greater than 50%. The occupancies of the strongest hydrogen bond (the one with the largest occupancy) do not change much for all such inhibitors except TKI258. This can be attributed to the absence of the 2,4-dimethoxy phenyl group in TKI258 unlike other inhibitors (refer to Figure 1) which provides enough empty space in the binding pocket for it to drift away from the protein.

The ΔΔG predictions for the complexes of TKI258 in case of remote mutants also become less accurate on “heating” as compared to standard TIES. In order to understand this, we accumulated the normalized frequency distributions of H1 and H2 for these complexes too as shown in Figures 12 and 13. It is interesting to note that, in the case of both remote mutants, the distributions are different at the two end-points of the standard TIES calculations such that there are additional small peaks in H1 and/or H2 distributions around 4 Å at the λ = 1 end-point. This may be related to the negative ΔΔG_calc^TIES predictions for both complexes of TKI258 (refer to Table 1 and Figure 4). However, in the case of λ-REST2 simulations, the situation appears to be even worse than the V555M–TKI258 complex with H1/H2 values reaching as high as 12 and 18 Å for I538V–TKI258 and N540S–TKI258 complexes, respectively. Both these complexes have peaks centered around 8 Å in the case of λ-REST2 simulations which correspond to the flipped conformation of the inhibitor when the fluorine atom of the TKI258 faces the hinge region of the protein while the H and O atoms involved in H1 and H2 point toward the opposite side (refer to Figure 10). This is further confirmed by the hydrogen bond analysis where the starred occupancies correspond to the hydrogen bonds formed by the inhibitor atoms facing opposite to the H and O involved in H1 and H2 (refer to Table S7 in the Supporting Information). Such flipped conformations correspond to a higher energy minimum which is inaccessible using standard MD simulations (and probably unobserved in experiments) but, due to the “heating” and exchanging of conformations between neighboring REST2-replicas, are observed in λ-REST2 simulations. This explains why the ΔΔG predictions using λ-REST2 become less accurate as compared to standard TIES. It should be noted that no such flipped conformations are observed in other inhibitors. This is partly due to the absence of the 2,4-dimethoxy phenyl group in the case of TKI258, which leaves an empty space in the binding pocket, and partly to relatively weaker interactions between TKI258 and the hinge region of the protein as compared to other inhibitors due to inclusion of an extra residue in the “hot” region (refer to Table S7 of the Supporting Information).

From the above discussion, it can be concluded that blind application of the λ-REST2 technique with the hope to improve sampling is not wise and may lead to potentially “unwanted” conformations. This can be further understood by considering a hypothetical situation where there are two potential energy minima separated by a barrier such that the lower energy minimum corresponds to the “wanted” conformation while the higher energy minimum corresponds to the “unwanted” conformation. During a λ-REST2 simulation, the intermediate λ-windows are “hot” and will sample both minima as well as the peak of the energy barrier. Although enhanced sampling is preferred in many cases, correctly predicting a physical observable of interest requires not only sufficient representative conformational states but the corresponding weights that describe the likelihoods of individual states. The latter are usually calculated based on the total energy of a system while the energy distributions of different states can be highly overlapping. It is therefore difficult to assess the relative likelihood of a state and its contribution to the prediction of the observable. Thus, while λ-REST2 can be a valuable technique, it should be used with care. Constraints from experiments, wherever available, should be used to infer the relevance of conformational space sampled by an enhanced MD simulation. In cases where there are no experimental data at all, many binding poses may need to be generated and evaluated, by methods such as docking and enhanced MD simulation, and the most plausible poses should be chosen based on their ranking scores. There are also data driven approaches such as random forests (31) and state-of-the-art neural network methods (32) which show some promise in this respect. As opposed to theory-led approaches, these data-driven machine learning approaches have a general limitation in biomedical research, (33) and their accuracies usually depend on the size and quality of training sets.

5. Conclusions

ARTICLE SECTIONS

Jump To

In this article, we describe the applications of ensemble-based approaches, with or without enhanced sampling protocols, to predict relative binding free energies of inhibitors to wild-type and mutant proteins. These approaches have been shown to be accurate and precise with effective control of errors for a range of target proteins and ligands. (11) Two challenging cases are investigated in the current study: one is a protein mutation within the binding site, which induces a large conformational change within one of the inhibitors; (25) another is the protein mutations remote from the binding site which do not have significant impact on the stability of the protein yet have an influence on inhibitor binding. (16)

The calculation of free energy changes caused by local mutations can benefit from enhanced sampling techniques such as REST2. The correct conformations are sampled in TIES-λ-REST2 simulations for the inhibitor AZD4547: while only one comformation is sampled for the inhibitor complexed with wild-type FGFR3, two conformations emerge when the gatekeeper residue is mutated from valine to methionine (V555M). Without enhanced sampling, the inhibitor remains trapped in its initial conformations, making the predicted free energy changes either overestimated or underestimated. TIES-λ-REST2, on the other hand, correctly predicts the free energy changes regardless of what initial conformation is used. As in our previous work, (11) the free energy estimator, MBAR, only offers a marginal improvement in the precisions of predictions but does not affect their accuracy.

For local mutations, the choice of a “hot” region is important in determining the efficiency and convergence of the free-energy calculations in simulations with REST2 approach. If the region is too small, the functionally relevant conformational space may not be explored sufficiently; while if it is too large, the system may drift away from those conformations leading to deteriorated predictions. Therefore, one requires some preliminary knowledge of the topological and physical properties of the protein–ligand systems for selection of an appropriate REST2 region. We do not question the utility of classical atomistic MD as a predictive tool for biomolecular systems, as many studies have proven the predictive ability of the method, including our two collaborative studies with pharmaceutical companies, (12,13) which were performed, initially blind, to investigate the binding affinities of compounds to protein targets. Our current study serves as a caution against the blind application of enhanced sampling approaches.

For remote mutations, however, the TIES-λ-REST2 approach does not generally improve the binding free-energy predictions. This is not surprising given that the mutations are far away from the bound inhibitors and affect the binding through an allosteric mechanism. Allostery involves conformational changes on length and/or time scales that are greater than standard atomistic molecular simulations can access. They only sample local conformational changes on a nanosecond time scale, which are not affected by remote mutations.

Supporting Information

ARTICLE SECTIONS

Jump To

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jctc.8b01118.

Tables containing individual ΔG_alch values for all the relative free energy calculations, and a table for the comparison of occupancies of hydrogen bonds between the inhibitors and the protein (PDF)
Topology and coordinate files for all ligand–protein complexes (ZIP)

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

ARTICLE SECTIONS

Jump To

Corresponding Author
- Peter V. Coveney - Centre for Computational Science, Department of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, United Kingdom; http://orcid.org/0000-0002-8787-7256; Email: [email protected]
Authors
- Agastya P. Bhati - Centre for Computational Science, Department of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, United Kingdom
- Shunzhou Wan - Centre for Computational Science, Department of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, United Kingdom; http://orcid.org/0000-0001-7192-1999
Funding
The authors would like to acknowledge the support of the MRC Medical Bioinformatics project (MR/L016311/1), the Qatar National Research Fund (7-1083-1-191), the EU H2020 projects ComPat (http://www.compat-project.eu/, Grant No. 671564), and CompBioMed (http://www.compbiomed.eu/, Grant No. 675451), NSF Award (https://www.nsf.gov/pubs/2017/nsf17542/nsf17542.htm, Award No. NSF 1713749) and funding from the UCL Provost. We made use of the BlueWaters supercomputer at the National Center for Supercomputing Applications of the University of Illinois at Urbana–Champaign (https://bluewaters.ncsa.illinois.edu), access to which was made available through the aforementioned NSF award, and the Titan supercomputer at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. A.P.B. was supported by an Overseas Research Scholarship from UCL and an Inlaks Scholarship from the Inlaks Shivdasani Foundation.
Notes
The authors declare no competing financial interest.

Acknowledgments

ARTICLE SECTIONS

Jump To

We thank our colleague Dr. David W. Wright for useful discussions.

References

ARTICLE SECTIONS

Jump To

This article references 33 other publications.

1
Wright, D. W.; Wan, S.; Shublaq, N.; Zasada, S. J.; Coveney, P. V. From base pair to bedside: molecular simulation and the translation of genomics to personalized medicine. WIREs Syst. Biol. Med. 2012, 4, 585– 598, DOI: 10.1002/wsbm.1186
Google Scholar
1
From base pair to bedside: molecular simulation and the translation of genomics to personalized medicine
Wright, David W.; Wan, Shunzhou; Shublaq, Nour; Zasada, Stefan J.; Coveney, Peter V.
Wiley Interdisciplinary Reviews: Systems Biology and Medicine (2012), 4 (6), 585-598CODEN: WIRSBW; ISSN:1939-005X. (Wiley-Blackwell)
A review. Despite the promises made that genomic sequencing would transform therapy by introducing a new era of personalized medicine, relatively few tangible breakthroughs have been made. This has led to the recognition that complex interactions at multiple spatial, temporal, and organizational levels may often combine to produce disease. Understanding this complexity requires that existing and future models are used and interpreted within a framework that incorporates knowledge derived from investigations at multiple levels of biol. function. It also requires a computational infrastructure capable of dealing with the vast quantities of data generated by genomic approaches. In this review, we discuss the use of mol. modeling to generate quant. and qual. insights at the smallest scales of the systems biol. hierarchy, how it can play an important role in the development of a systems understanding of disease and in the application of such knowledge to help discover new therapies and target existing ones on a personal level.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3sXpslSgsg%253D%253D&md5=b6bf5cb6b9fab33543986b9f8ceed6ec
2
Wang, L. Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field. J. Am. Chem. Soc. 2015, 137, 2695– 2703, DOI: 10.1021/ja512751q
Google Scholar
2
Accurate and Reliable Prediction of Relative Ligand Binding Potency in Prospective Drug Discovery by Way of a Modern Free-Energy Calculation Protocol and Force Field
Wang, Lingle; Wu, Yujie; Deng, Yuqing; Kim, Byungchan; Pierce, Levi; Krilov, Goran; Lupyan, Dmitry; Robinson, Shaughnessy; Dahlgren, Markus K.; Greenwood, Jeremy; Romero, Donna L.; Masse, Craig; Knight, Jennifer L.; Steinbrecher, Thomas; Beuming, Thijs; Damm, Wolfgang; Harder, Ed; Sherman, Woody; Brewer, Mark; Wester, Ron; Murcko, Mark; Frye, Leah; Farid, Ramy; Lin, Teng; Mobley, David L.; Jorgensen, William L.; Berne, Bruce J.; Friesner, Richard A.; Abel, Robert
Journal of the American Chemical Society (2015), 137 (7), 2695-2703CODEN: JACSAT; ISSN:0002-7863. (American Chemical Society)
Designing tight-binding ligands is a primary objective of small-mol. drug discovery. Over the past few decades, free-energy calcns. have benefited from improved force fields and sampling algorithms, as well as the advent of low-cost parallel computing. However, it has proven to be challenging to reliably achieve the level of accuracy that would be needed to guide lead optimization (∼5× in binding affinity) for a wide range of ligands and protein targets. Not surprisingly, widespread com. application of free-energy simulations has been limited due to the lack of large-scale validation coupled with the tech. challenges traditionally assocd. with running these types of calcns. Here, we report an approach that achieves an unprecedented level of accuracy across a broad range of target classes and ligands, with retrospective results encompassing 200 ligands and a wide variety of chem. perturbations, many of which involve significant changes in ligand chem. structures. In addn., we have applied the method in prospective drug discovery projects and found a significant improvement in the quality of the compds. synthesized that have been predicted to be potent. Compds. predicted to be potent by this approach have a substantial redn. in false positives relative to compds. synthesized on the basis of other computational or medicinal chem. approaches. Furthermore, the results are consistent with those obtained from our retrospective studies, demonstrating the robustness and broad range of applicability of this approach, which can be used to drive decisions in lead optimization.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhsF2iuro%253D&md5=37a4f4a6c085f47ed531342643b6c33b
3
Bernardi, R. C.; Melo, M. C.; Schulten, K. Enhanced sampling techniques in molecular dynamics simulations of biological systems. Biochim. Biophys. Acta, Gen. Subj. 2015, 1850, 872– 877, DOI: 10.1016/j.bbagen.2014.10.019
Google Scholar
3
Enhanced sampling techniques in molecular dynamics simulations of biological systems
Bernardi, Rafael C.; Melo, Marcelo C. R.; Schulten, Klaus
Biochimica et Biophysica Acta, General Subjects (2015), 1850 (5), 872-877CODEN: BBGSB3; ISSN:0304-4165. (Elsevier B.V.)
A review. Mol. dynamics has emerged as an important research methodol. covering systems to the level of millions of atoms. However, insufficient sampling often limits its application. The limitation is due to rough energy landscapes, with many local min. sepd. by high-energy barriers, which govern the biomol. motion. In the past few decades methods have been developed that address the sampling problem, such as replica-exchange mol. dynamics, metadynamics and simulated annealing. Here the authors present an overview over theses sampling methods in an attempt to shed light on which should be selected depending on the type of system property studied. Enhanced sampling methods have been employed for a broad range of biol. systems and the choice of a suitable method is connected to biol. and phys. characteristics of the system, in particular system size. While metadynamics and replica-exchange mol. dynamics are the most adopted sampling methods to study biomol. dynamics, simulated annealing is well suited to characterize very flexible systems. The use of annealing methods for a long time was restricted to simulation of small proteins; however, a variant of the method, generalized simulated annealing, can be employed at a relatively low computational cost to large macromol. complexes. Mol. dynamics trajectories frequently do not reach all relevant conformational substates, for example those connected with biol. function, a problem that can be addressed by employing enhanced sampling algorithms. This article is part of a Special Issue entitled Recent developments of mol. dynamics.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2cXhvVanurbN&md5=44e3ac1aa042c35ee08aa86a6a63e78e
4
Valsson, O.; Parrinello, M. Variational approach to enhanced sampling and free energy calculations. Phys. Rev. Lett. 2014, 113, 090601, DOI: 10.1103/PhysRevLett.113.090601
Google Scholar
4
Variational approach to enhanced sampling and free energy calculations
Valsson, Omar; Parrinello, Michele
Physical Review Letters (2014), 113 (9), 090601CODEN: PRLTAO; ISSN:0031-9007. (American Physical Society)
The ability of widely used sampling methods, such as mol. dynamics or Monte Carlo simulations, to explore complex free energy landscapes is severely hampered by the presence of kinetic bottlenecks. A large no. of solns. have been proposed to alleviate this problem. Many are based on the introduction of a bias potential which is a function of a small no. of collective variables. However constructing such a bias is not simple. Here we introduce a functional of the bias potential and an assocd. variational principle. The bias that minimizes the functional relates in a simple way to the free energy surface. This variational principle can be turned into a practical, efficient, and flexible sampling method. A no. of numerical examples are presented which include the detn. of a three-dimensional free energy surface. We argue that, beside being numerically advantageous, our variational approach provides a convenient and novel standpoint for looking at the sampling problem.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2cXhslShtL7O&md5=4d66ba6f0d445e693e28dbeae09cd936
5
Fukunishi, H.; Watanabe, O.; Takada, S. On the Hamiltonian replica exchange method for efficient sampling of biomolecular systems: Application to protein structure prediction. J. Chem. Phys. 2002, 116, 9058– 9067, DOI: 10.1063/1.1472510
Google Scholar
5
On the Hamiltonian replica exchange method for efficient sampling of biomolecular systems: Application to protein structure prediction
Fukunishi, Hiroaki; Watanabe, Osamu; Takada, Shoji
Journal of Chemical Physics (2002), 116 (20), 9058-9067CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
Motivated by the protein structure prediction problem, we develop two variants of the Hamiltonian replica exchange methods (REMs) for efficient configuration sampling, (1) the scaled hydrophobicity REM and (2) the phantom chain REM, and compare their performance with the ordinary REM. We first point out that the ordinary REM has a shortage for the application to large systems such as biomols. and that the Hamiltonian REM, an alternative formulation of the REM, can give a remedy for it. We then propose two examples of the Hamiltonian REM that are suitable for a coarse-grained protein model. (1) The scaled hydrophobicity REM preps. replicas that are characterized by various strengths of hydrophobic interaction. The strongest interaction that mimics aq. soln. environment makes proteins folding, while weakened hydrophobicity unfolds proteins as in org. solvent. Exchange between these environments enables proteins to escape from misfolded traps and accelerate conformational search. This resembles the roles of mol. chaperone that assist proteins to fold in vivo. (2) The phantom chain REM uses replicas that allow various degrees of at. overlaps. By allowing at. overlap in some of replicas, the peptide chain can cross over itself, which can accelerate conformation sampling. Using a coarse-gained model we developed, we compute equil. probability distributions for poly-alanine 16-mer and for a small protein by these REMs and compare the accuracy of the results. We see that the scaled hydrophobicity REM is the most efficient method among the three REMs studied.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD38XjsFKmsLo%253D&md5=7ac571a5afdd63b0b4b29cfdec06f53b
6
Wang, L.; Friesner, R. A.; Berne, B. J. Replica exchange with solute scaling: A more efficient version of replica exchange with solute tempering (REST2). J. Phys. Chem. B 2011, 115, 9431– 9438, DOI: 10.1021/jp204407d
Google Scholar
6
Replica Exchange with Solute Scaling: A More Efficient Version of Replica Exchange with Solute Tempering (REST2)
Wang, Lingle; Friesner, Richard A.; Berne, B. J.
Journal of Physical Chemistry B (2011), 115 (30), 9431-9438CODEN: JPCBFK; ISSN:1520-5207. (American Chemical Society)
A small change in the Hamiltonian scaling in Replica Exchange with Solute Tempering (REST) is found to improve its sampling efficiency greatly, esp. for the sampling of aq. protein solns. in which there are large-scale solute conformation changes. Like the original REST (REST1), the new version (which the authors call REST2) also bypasses the poor scaling with system size of the std. Temp. Replica Exchange Method (TREM), reducing the no. of replicas (parallel processes) from what must be used in TREM. This redn. is accomplished by deforming the Hamiltonian function for each replica in such a way that the acceptance probability for the exchange of replica configurations does not depend on the no. of explicit water mols. in the system. For proof of concept, REST2 is compared with TREM and with REST1 for the folding of the trpcage and β-hairpin in water. The comparisons confirm that REST2 greatly reduces the no. of CPUs required by regular replica exchange and greatly increases the sampling efficiency over REST1. This method reduces the CPU time required for calcg. thermodn. avs. and for the ab initio folding of proteins in explicit water.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXosFymurs%253D&md5=8acb0d670bcb70eacfc30e32d5dfb6dd
7
Wang, L.; Berne, B. J.; Friesner, R. A. On achieving high accuracy and reliability in the calculation of relative protein–ligand binding affinities. Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 1937– 1942, DOI: 10.1073/pnas.1114017109
Google Scholar
7
On achieving high accuracy and reliability in the calculation of relative protein-ligand binding affinities
Wang, Lingle; Berne, B. J.; Friesner, Richard A.
Proceedings of the National Academy of Sciences of the United States of America (2012), 109 (6), 1937-1942, S1937/1-S1937/7CODEN: PNASA6; ISSN:0027-8424. (National Academy of Sciences)
We apply a free energy perturbation simulation method, free energy perturbation/replica exchange with solute tempering, to two modifications of protein-ligand complexes that lead to significant conformational changes, the first in the protein and the second in the ligand. The approach is shown to facilitate sampling in these challenging cases where high free energy barriers sep. the initial and final conformations and leads to superior convergence of the free energy as demonstrated both by consistency of the results (independence from the starting conformation) and agreement with exptl. binding affinity data. The second case, consisting of two neutral thrombin ligands that are taken from a recent medicinal chem. program for this interesting pharmaceutical target, is of particular significance in that it demonstrates that good results can be obtained for large, complex ligands, as opposed to relatively simple model systems. To achieve quant. agreement with expt. in the thrombin case, a next generation force field, Optimized Potentials for Liq. Simulations 2.0, is required, which provides superior charges and torsional parameters as compared to earlier alternatives.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC38XivVSht74%253D&md5=23db7d18e624d5e1214bbd6702619a87
8
Paliwal, H.; Shirts, M. R. A benchmark test set for alchemical free energy transformations and its use to quantify error in common free energy methods. J. Chem. Theory Comput. 2011, 7, 4115– 4134, DOI: 10.1021/ct2003995
Google Scholar
8
A Benchmark Test Set for Alchemical Free Energy Transformations and Its Use to Quantify Error in Common Free Energy Methods
Paliwal, Himanshu; Shirts, Michael R.
Journal of Chemical Theory and Computation (2011), 7 (12), 4115-4134CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
There is a significant need for improved tools to validate thermophys. quantities computed via mol. simulation. In this paper we present the initial version of a benchmark set of testing methods for calcg. free energies of mol. transformation in soln. This set is based on mol. changes common to many mol. design problems, such as insertion and deletion of at. sites and changing at. partial charges. We use this benchmark set to compare the statistical efficiency, reliability, and quality of uncertainty ests. for a no. of published free energy methods, including thermodn. integration, free energy perturbation, the Bennett acceptance ratio (BAR) and its multistate equiv. MBAR. We identify MBAR as the consistently best performing method, though other methods are frequently comparable in reliability and accuracy in many cases. We demonstrate that assumptions of Gaussian distributed errors in free energies are usually valid for most methods studied. We demonstrate that bootstrap error estn. is a robust and useful technique for estg. statistical variance for all free energy methods studied. This benchmark set is provided in a no. of different file formats with the hope of becoming a useful and general tool for method comparisons.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXhtlaqtrvM&md5=41383373300216165a7f1724ad535451
9
Shirts, M. R.; Chodera, J. D. Statistically optimal analysis of samples from multiple equilibrium states. J. Chem. Phys. 2008, 129, 124105, DOI: 10.1063/1.2978177
Google Scholar
9
Statistically optimal analysis of samples from multiple equilibrium states
Shirts, Michael R.; Chodera, John D.
Journal of Chemical Physics (2008), 129 (12), 124105/1-124105/10CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
We present a new estimator for computing free energy differences and thermodn. expectations as well as their uncertainties from samples obtained from multiple equil. states via either simulation or expt. The estimator, which we call the multistate Bennett acceptance ratio estimator (MBAR) because it reduces to the Bennett acceptance ratio estimator (BAR) when only two states are considered, has significant advantages over multiple histogram reweighting methods for combining data from multiple states. It does not require the sampled energy range to be discretized to produce histograms, eliminating bias due to energy binning and significantly reducing the time complexity of computing a soln. to the estg. equations in many cases. Addnl., an est. of the statistical uncertainty is provided for all estd. quantities. In the large sample limit, MBAR is unbiased and has the lowest variance of any known estimator for making use of equil. data collected from multiple states. We illustrate this method by producing a highly precise est. of the potential of mean force for a DNA hairpin system, combining data from multiple optical tweezer measurements under const. force bias. (c) 2008 American Institute of Physics.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXht1WnsL7F&md5=479183e1f45fc58dd7c6e5ef1e73d45d
10
Bhati, A. P.; Wan, S.; Wright, D. W.; Coveney, P. V. Rapid, accurate, precise, and reliable relative free energy prediction using ensemble based thermodynamic integration. J. Chem. Theory Comput. 2017, 13, 210– 222, DOI: 10.1021/acs.jctc.6b00979
Google Scholar
10
Rapid, Accurate, Precise, and Reliable Relative Free Energy Prediction Using Ensemble Based Thermodynamic Integration
Bhati, Agastya P.; Wan, Shunzhou; Wright, David W.; Coveney, Peter V.
Journal of Chemical Theory and Computation (2017), 13 (1), 210-222CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
The accurate prediction of the binding affinities of ligands to proteins is a major goal in drug discovery and personalized medicine. The time taken to make such predictions is of similar importance to their accuracy, precision and reliability. In the last few years, an ensemble based mol. dynamics approach has been proposed that provides a route to reliable predictions of free energies based on the mol. mechanics Poisson-Boltzmann surface area method which meets the requirements of accuracy, precision and reliability. Here, we describe an equiv. methodol. based on thermodn. integration to substantially improve the accuracy, precision and reliability of calcd. relative binding free energies. We report the performance of the method when applied to a diverse set of protein targets and ligands. The results are in very good agreement with exptl. data (90% of calcns. agree to within 1 kcal/mol) while the method is reproducible by construction. Statistical uncertainties of the order of 0.5 kcal/mol or less are achieved. We present a systematic account of how the uncertainty in the predictions may be estd.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitVGntrrJ&md5=510b70188112a3578030e291ce19b127
11
Bhati, A. P.; Wan, S.; Hu, Y.; Sherborne, B.; Coveney, P. V. Uncertainty quantification in alchemical free energy methods. J. Chem. Theory Comput. 2018, 14, 2867– 2880, DOI: 10.1021/acs.jctc.7b01143
Google Scholar
11
Uncertainty Quantification in Alchemical Free Energy Methods
Bhati, Agastya P.; Wan, Shunzhou; Hu, Yuan; Sherborne, Brad; Coveney, Peter V.
Journal of Chemical Theory and Computation (2018), 14 (6), 2867-2880CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
Alchem. free energy methods have gained much importance recently from several reports of improved ligand-protein binding affinity predictions based on their implementation using mol. dynamics simulations. A large no. of variants of such methods implementing different accelerated sampling techniques and free energy estimators are available, each claimed to be better than the others in its own way. However, the key features of reproducibility and quantification of assocd. uncertainties in such methods have barely been discussed. Here, we apply a systematic protocol for uncertainty quantification to a no. of popular alchem. free energy methods, covering both abs. and relative free energy predictions. We show that a reliable measure of error estn. is provided by ensemble simulation-an ensemble of independent MD simulations-which applies irresp. of the free energy method. The need to use ensemble methods is fundamental and holds regardless of the duration of time of the mol. dynamics simulations performed.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXotFaqtr4%253D&md5=5f80e23df9c8aafb58914af06de24e1f
12
Wan, S.; Bhati, A. P.; Skerratt, S.; Omoto, K.; Shanmugasundaram, V.; Bagal, S. K.; Coveney, P. V. Evaluation and characterization of Trk kinase inhibitors for the treatment of pain: reliable binding affinity predictions from theory and computation. J. Chem. Inf. Model. 2017, 57, 897– 909, DOI: 10.1021/acs.jcim.6b00780
Google Scholar
12
Evaluation and Characterization of Trk Kinase Inhibitors for the Treatment of Pain: Reliable Binding Affinity Predictions from Theory and Computation
Wan, Shunzhou; Bhati, Agastya P.; Skerratt, Sarah; Omoto, Kiyoyuki; Shanmugasundaram, Veerabahu; Bagal, Sharan K.; Coveney, Peter V.
Journal of Chemical Information and Modeling (2017), 57 (4), 897-909CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
Optimization of ligand binding affinity to the target protein of interest is a primary objective in small-mol. drug discovery. Until now, the prediction of binding affinities by computational methods has not been widely applied in the drug discovery process, mainly due to its lack of accuracy and reproducibility, as well as the long turnaround times required to obtain results. Herein, the authors report on a collaborative study that compares tropomyosin receptor kinase A (TrkA) binding affinity predictions using two recently formulated fast computational approaches - namely ESMACS (Enhanced Sampling of Mol. dynamics with Approxn. of Continuum Solvent) and TIES (Thermodn. Integration with Enhanced Sampling) - to exptl. derived TrkA binding affinities for a set of Pfizer pan-Trk compds. ESMACS gives precise and reproducible results and is applicable to highly diverse sets of compds. It also provides detailed chem. insight into the nature of ligand-protein binding. TIES can predict and thus optimize more subtle changes in binding affinities between compds. of similar structure. Individual binding affinities were calcd. in a few hours, exhibiting good correlations with the exptl. data of 0.79 and 0.88 from ESMACS and TIES approaches resp. The speed, level of accuracy and precision of the calcns. are such that the affinity predictions can be used to rapidly explain the effects of compd. modifications on TrkA binding affinity. The methods could therefore be used as tools to guide lead optimization efforts across multiple prospective structurally-enabled programs in the drug discovery setting for a wide range of compds. and targets.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXksVyrurc%253D&md5=2e366f56c1ee8c2f2b2760b9098e2030
13
Wan, S.; Bhati, A. P.; Zasada, S. J.; Wall, I.; Green, D.; Bamborough, P.; Coveney, P. V. Rapid and reliable binding affinity prediction of bromodomain inhibitors: A computational study. J. Chem. Theory Comput. 2017, 13, 784– 795, DOI: 10.1021/acs.jctc.6b00794
Google Scholar
13
Rapid and Reliable Binding Affinity Prediction of Bromodomain Inhibitors: A Computational Study
Wan, Shunzhou; Bhati, Agastya P.; Zasada, Stefan J.; Wall, Ian; Green, Darren; Bamborough, Paul; Coveney, Peter V.
Journal of Chemical Theory and Computation (2017), 13 (2), 784-795CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
Binding free energies of bromodomain inhibitors are calcd. with recently formulated approaches, namely ESMACS (enhanced sampling of mol. dynamics with approxn. of continuum solvent) and TIES (thermodn. integration with enhanced sampling). A set of compds. is provided by GlaxoSmithKline, which represents a range of chem. functionality and binding affinities. The predicted binding free energies exhibit a good Spearman correlation of 0.78 with the exptl. data from the 3-trajectory ESMACS, and an excellent correlation of 0.92 from the TIES approach where applicable. Given access to suitable high end computing resources and a high degree of automation, the authors can compute individual binding affinities in a few hours with precisions no greater than 0.2 kcal/mol for TIES, and no larger than 0.34 kcal/mol and 1.71 kcal/mol for the 1- and 3-trajectory ESMACS approaches.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitFektLnF&md5=713436084662420482684fb50db0832e
14
Porta, R.; Borea, R.; Coelho, A.; Khan, S.; Araújo, A.; Reclusa, P.; Franchina, T.; Steen, N. V. D.; Dam, P. V.; Ferri, J.; Sirera, R.; Naing, A.; Hong, D.; Rolfo, C. FGFR a promising druggable target in cancer: Molecular biology and new drugs. Crit. Rev. Oncol. Hematol. 2017, 113, 256– 267, DOI: 10.1016/j.critrevonc.2017.02.018
Google Scholar
14
FGFR a promising druggable target in cancer: Molecular biology and new drugs
Porta Rut; Borea Roberto; Coelho Andreia; Reclusa Pablo; Van Dam Peter; Ferri Jose; Sirera Rafael; Khan Shahanavaj; Araujo Antonio; Franchina Tindara; Van Der Steen Nele; Naing Aung; Hong David; Rolfo Christian
Critical reviews in oncology/hematology (2017), 113 (), 256-267 ISSN:.
INTRODUCTION: The Fibroblast Growth Factor Receptor (FGFR) family consists of Tyrosine Kinase Receptors (TKR) involved in several biological functions. Recently, alterations of FGFR have been reported to be important for progression and development of several cancers. In this setting, different studies are trying to evaluate the efficacy of different therapies targeting FGFR. AREAS COVERED: This review summarizes the current status of treatments targeting FGFR, focusing on the trials that are evaluating the FGFR profile as inclusion criteria: Multi-Target, Pan-FGFR Inhibitors and anti-FGF (Fibroblast Growth Factor)/FGFR Monoclonal Antibodies. EXPERT OPINION: Most of the TKR share intracellular signaling pathways; therefore, cancer cells tend to overcome the inhibition of one tyrosine kinase receptor by activating another. The future of TKI (Tyrosine Kinase Inhibitor) therapy will potentially come from multi-targeted TKIs that target different TKR simultaneously. It is crucial to understand the interaction of the FGF-FGFR axis with other known driver TKRs. Based on this, it is possible to develop therapeutic strategies targeting multiple connected TKRs at once. One correct step in this direction is the reassessment of multi target inhibitors considering the FGFR status of the tumor. Another opportunity arises from assessing the use of FGFR TKI on patients harboring FGFR alterations.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1cvps1ajsA%253D%253D&md5=d03086e258264a12c8c5cc7f2c5de8e3
15
Bunney, T. D.; Wan, S.; Thiyagarajan, N.; Sutto, L.; Williams, S. V.; Ashford, P.; Koss, H.; Knowles, M. a.; Gervasio, F. L.; Coveney, P. V.; Katan, M. The effect of mutations on drug sensitivity and kinase activity of fibroblast growth factor receptors: A combined experimental and theoretical study. EBioMedicine 2015, 2, 194– 204, DOI: 10.1016/j.ebiom.2015.02.009
Google Scholar
15
The Effect of Mutations on Drug Sensitivity and Kinase Activity of Fibroblast Growth Factor Receptors: A Combined Experimental and Theoretical Study
Bunney Tom D; Thiyagarajan Nethaji; Ashford Paul; Katan Matilda; Wan Shunzhou; Coveney Peter V; Sutto Ludovico; Gervasio Francesco L; Williams Sarah V; Knowles Margaret A; Koss Hans
EBioMedicine (2015), 2 (3), 194-204 ISSN:.
Fibroblast growth factor receptors (FGFRs) are recognized therapeutic targets in cancer. We here describe insights underpinning the impact of mutations on FGFR1 and FGFR3 kinase activity and drug efficacy, using a combination of computational calculations and experimental approaches including cellular studies, X-ray crystallography and biophysical and biochemical measurements. Our findings reveal that some of the tested compounds, in particular TKI258, could provide therapeutic opportunity not only for patients with primary alterations in FGFR but also for acquired resistance due to the gatekeeper mutation. The accuracy of the computational methodologies applied here shows a potential for their wider application in studies of drug binding and in assessments of functional and mechanistic impacts of mutations, thus assisting efforts in precision medicine.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC2srotlKmtA%253D%253D&md5=8414f3530d24bbba0ea8cc98cb784b96
16
Patani, H.; Bunney, T. D.; Thiyagarajan, N.; Norman, R. A.; Ogg, D.; Breed, J.; Ashford, P.; Potterton, A.; Edwards, M.; Katan, M. Landscape of activating cancer mutations in FGFR kinases and their differential responses to inhibitors in clinical use. Oncotarget 2016, 7, 24252– 24268, DOI: 10.18632/oncotarget.8132
Google Scholar
16
Landscape of activating cancer mutations in FGFR kinases and their differential responses to inhibitors in clinical use
Patani Harshnira; Bunney Tom D; Thiyagarajan Nethaji; Ashford Paul; Potterton Andrew; Edwards Mina; Pang Camilla S M; Orengo Christine; Katan Matilda; Norman Richard A; Ogg Derek; Breed Jason; Phillips Chris; Williams Sarah V; Knowles Margaret A; Thomson Gary S; Breeze Alexander L
Oncotarget (2016), 7 (17), 24252-68 ISSN:.
Frequent genetic alterations discovered in FGFRs and evidence implicating some as drivers in diverse tumors has been accompanied by rapid progress in targeting FGFRs for anticancer treatments. Wider assessment of the impact of genetic changes on the activation state and drug responses is needed to better link the genomic data and treatment options. We here apply a direct comparative and comprehensive analysis of FGFR3 kinase domain variants representing the diversity of point-mutations reported in this domain. We reinforce the importance of N540K and K650E and establish that not all highly activating mutations (for example R669G) occur at high-frequency and conversely, that some "hotspots" may not be linked to activation. Further structural characterization consolidates a mechanistic view of FGFR kinase activation and extends insights into drug binding. Importantly, using several inhibitors of particular clinical interest (AZD4547, BGJ-398, TKI258, JNJ42756493 and AP24534), we find that some activating mutations (including different replacements of the same residue) result in distinct changes in their efficacy. Considering that there is no approved inhibitor for anticancer treatments based on FGFR-targeting, this information will be immediately translatable to ongoing clinical trials.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC28fhsFShsw%253D%253D&md5=beb65006a52ef6062bab9c9d8086bf1d
17
Azam, M.; Seeliger, M. a.; Gray, N. S.; Kuriyan, J.; Daley, G. Q. Activation of tyrosine kinases by mutation of the gatekeeper threonine. Nat. Struct. Mol. Biol. 2008, 15, 1109– 1118, DOI: 10.1038/nsmb.1486
Google Scholar
17
Activation of tyrosine kinases by mutation of the gatekeeper threonine
Azam, Mohammad; Seeliger, Markus A.; Gray, Nathanael S.; Kuriyan, John; Daley, George Q.
Nature Structural & Molecular Biology (2008), 15 (10), 1109-1118CODEN: NSMBCU; ISSN:1545-9993. (Nature Publishing Group)
Protein kinases targeted by small-mol. inhibitors develop resistance through mutation of the 'gatekeeper' threonine residue of the active site. Here we show that the gatekeeper mutation in the cellular forms of c-ABL, c-SRC, platelet-derived growth factor receptor-α and -β, and epidermal growth factor receptor activates the kinase and promotes malignant transformation of BaF3 cells. Structural anal. reveals that a network of hydrophobic interactions - the hydrophobic spine - characteristic of the active kinase conformation is stabilized by the gatekeeper substitution. Substitution of glycine for the residues constituting the spine disrupts the hydrophobic connectivity and inactivates the kinase. Furthermore, a small-mol. inhibitor (compd. 14) that maximizes complementarity with the dismantled spine inhibits the gatekeeper mutation of BCR-ABL-T315I. These results demonstrate that mutation of the gatekeeper threonine is a common mechanism of activation for tyrosine kinases and provide structural insights to guide the development of next-generation inhibitors.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXht1Sisb%252FM&md5=4afc4e5a0850a4510ced3fa5fbda97fa
18
van Gunsteren, W. F.; Daura, X.; Hansen, N.; Mark, A. E.; Oostenbrink, C.; Riniker, S.; Smith, L. J. Validation of Molecular Simulation: An Overview of Issues. Angew. Chem., Int. Ed. 2018, 57, 884– 902, DOI: 10.1002/anie.201702945
Google Scholar
18
Validation of Molecular Simulation: An Overview of Issues
van Gunsteren, Wilfred F.; Daura, Xavier; Hansen, Niels; Mark, Alan E.; Oostenbrink, Chris; Riniker, Sereina; Smith, Lorna J.
Angewandte Chemie, International Edition (2018), 57 (4), 884-902CODEN: ACIEF5; ISSN:1433-7851. (Wiley-VCH Verlag GmbH & Co. KGaA)
Computer simulation of mol. systems enables structure-energy-function relationships of mol. processes to be described at the sub-at., at., supra-at., or supra-mol. level. To interpret results of such simulations appropriately, the quality of the calcd. properties must be evaluated. This depends on the way the simulations are performed and on the way they are validated by comparison to values Qexp of exptl. observable quantities Q. One must consider (1) the accuracy of Qexp, (2) the accuracy of the function Q(rN) used to calc. a Q-value based on a mol. configuration rN of N particles, (3) the sensitivity of the function Q(rN) to the configuration rN, (4) the relative time scales of the simulation and expt., (5) the degree to which the calcd. and exptl. properties are equiv., and (6) the degree to which the system simulated matches the exptl. conditions. Exptl. data is limited in scope and generally corresponds to avs. over both time and space. A crit. anal. of the various factors influencing the apparent degree of (dis)agreement between simulations and expt. is presented and illustrated using examples from the literature. What can be done to enhance the validation of mol. simulation is also discussed.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhs1ertQ%253D%253D&md5=d38b0f3fd0054e0fab46023a2c597442
19
Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kal, L.; Schulten, K. Scalable molecular dynamics with NAMD. J. Comput. Chem. 2005, 26, 1781– 1802, DOI: 10.1002/jcc.20289
Google Scholar
19
Scalable molecular dynamics with NAMD
Phillips, James C.; Braun, Rosemary; Wang, Wei; Gumbart, James; Tajkhorshid, Emad; Villa, Elizabeth; Chipot, Christophe; Skeel, Robert D.; Kale, Laxmikant; Schulten, Klaus
Journal of Computational Chemistry (2005), 26 (16), 1781-1802CODEN: JCCHDD; ISSN:0192-8651. (John Wiley & Sons, Inc.)
NAMD is a parallel mol. dynamics code designed for high-performance simulation of large biomol. systems. NAMD scales to hundreds of processors on high-end parallel platforms, as well as tens of processors on low-cost commodity clusters, and also runs on individual desktop and laptop computers. NAMD works with AMBER and CHARMM potential functions, parameters, and file formats. This article, directed to novices as well as experts, first introduces concepts and methods used in the NAMD program, describing the classical mol. dynamics force field, equations of motion, and integration methods along with the efficient electrostatics evaluation algorithms employed and temp. and pressure controls used. Features for steering the simulation across barriers and for calcg. both alchem. and conformational free energy differences are presented. The motivations for and a roadmap to the internal design of NAMD, implemented in C++ and based on Charm++ parallel objects, are outlined. The factors affecting the serial and parallel performance of a simulation are discussed. Finally, typical NAMD use is illustrated with representative applications to a small, a medium, and a large biomol. system, highlighting particular features of NAMD, for example, the Tcl scripting language. The article also provides a list of the key features of NAMD and discusses the benefits of combining NAMD with the mol. graphics/sequence anal. software VMD and the grid computing/collab. software BioCoRE. NAMD is distributed free of charge with source code at www.ks.uiuc.edu.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2MXht1SlsbbJ&md5=189051128443b547f4300a1b8fb0e034
20
Coveney, P. V.; Wan, S. On the calculation of equilibrium thermodynamic properties from molecular dynamics. Phys. Chem. Chem. Phys. 2016, 18, 30236– 30240, DOI: 10.1039/C6CP02349E
Google Scholar
20
On the calculation of equilibrium thermodynamic properties from molecular dynamics
Coveney, Peter V.; Wan, Shunzhou
Physical Chemistry Chemical Physics (2016), 18 (44), 30236-30240CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
The purpose of statistical mechanics is to provide a route to the calcn. of macroscopic properties of matter from their constituent microscopic components. It is well known that the macrostates emerge as ensemble avs. of microstates. However, this is more often stated than implemented in computer simulation studies. Here we consider foundational aspects of statistical mechanics which are overlooked in most textbooks and research articles that purport to compute macroscopic behavior from microscopic descriptions based on classical mechanics and show how due attention to these issues leads in directions which have not been widely appreciated in the field of mol. dynamics simulation.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XntVyisr8%253D&md5=f7950af8255cf494a117d46bc1b5191e
21
Lim, N. M.; Wang, L.; Abel, R.; Mobley, D. L. Sensitivity in binding free energies due to protein reorganization. J. Chem. Theory Comput. 2016, 12, 4620– 4631, DOI: 10.1021/acs.jctc.6b00532
Google Scholar
21
Sensitivity in Binding Free Energies Due to Protein Reorganization
Lim, Nathan M.; Wang, Lingle; Abel, Robert; Mobley, David L.
Journal of Chemical Theory and Computation (2016), 12 (9), 4620-4631CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
Tremendous recent improvements in computer hardware, coupled with advances in sampling techniques and force fields, are now allowing protein-ligand binding free energy calcns. to be routinely used to aid pharmaceutical drug discovery projects. However, despite these recent innovations, there are still needs for further improvement in sampling algorithms to more adequately sample protein motion relevant to protein-ligand binding. Here, we report our work identifying and studying such clear and remaining needs in the apolar cavity of T4 lysozyme L99A. In this study, we model recent exptl. results that show the progressive opening of the binding pocket in response to a series of homologous ligands. Even while using enhanced sampling techniques, we demonstrate that the predicted relative binding free energies (RBFE) are sensitive to the initial protein conformational state. Particularly, we highlight the importance of sufficient sampling of protein conformational changes and demonstrate how inclusion of three key protein residues in the "hot" region of the FEP/REST simulation improves the sampling and resolves this sensitivity, given enough simulation time.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28Xht1CltrbO&md5=5f4525fb7a55b6c9ad0510168e13f316
22
Clark, A. J.; Gindin, T.; Zhang, B.; Wang, L.; Abel, R.; Murret, C. S.; Xu, F.; Bao, A.; Lu, N. J.; Zhou, T.; Kwong, P. D.; Shapiro, L.; Honig, B.; Friesner, R. A. Free energy perturbation calculation of relative binding free energy between broadly neutralizing antibodies and the gp120 glycoprotein of HIV-1. J. Mol. Biol. 2017, 429, 930– 947, DOI: 10.1016/j.jmb.2016.11.021
Google Scholar
22
Free energy perturbation calculation of relative binding free energy between broadly neutralizing antibodies and the gp120 glycoprotein of HIV-1
Clark, Anthony J.; Gindin, Tatyana; Zhang, Baoshan; Wang, Lingle; Abel, Robert; Murret, Colleen S.; Xu, Fang; Bao, Amy; Lu, Nina J.; Zhou, Tongqing; Kwong, Peter D.; Shapiro, Lawrence; Honig, Barry; Friesner, Richard A.
Journal of Molecular Biology (2017), 429 (7), 930-947CODEN: JMOBAK; ISSN:0022-2836. (Elsevier Ltd.)
Direct calcn. of relative binding affinities between antibodies and antigens is a long-sought goal. However, despite substantial efforts, no generally applicable computational method has been described. Here, we describe a systematic free energy perturbation (FEP) protocol and calc. the binding affinities between the gp120 envelope glycoprotein of HIV-1 and three broadly neutralizing antibodies (bNAbs) of the VRC01 class. The protocol has been adapted from successful studies of small mols. to address the challenges assocd. with modeling protein-protein interactions. Specifically, we built homol. models of the three antibody-gp120 complexes, extended the sampling times for large bulky residues, incorporated the modeling of glycans on the surface of gp120, and utilized continuum solvent-based loop prediction protocols to improve sampling. We present three exptl. surface plasmon resonance data sets, in which antibody residues in the antibody/gp120 interface were systematically mutated to alanine. The RMS error in the large set (55 total cases) of FEP tests as compared to these expts., 0.68 kcal/mol, is near exptl. accuracy, and it compares favorably with the results obtained from a simpler, empirical methodol. The correlation coeff. for the combined data set including residues with glycan contacts, R2 = 0.49, should be sufficient to guide the choice of residues for antibody optimization projects, assuming that this level of accuracy can be realized in prospective prediction. More generally, these results are encouraging with regard to the possibility of using an FEP approach to calc. the magnitude of protein-protein binding affinities.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitVSmsL3I&md5=ff264d642726916a45bae5b121bbdd06
23
Fratev, F.; Sirimulla, S. An improved free energy perturbation FEP+ sampling protocol for flexible ligand-binding domains. ChemRxiv 2018, DOI: 10.26434/chemrxiv.6204167.v1 .
Google Scholar
There is no corresponding record for this reference.
24
Fiser, A.; Sali, A. ModLoop: automated modeling of loops in protein structures. Bioinformatics 2003, 19, 2500– 2501, DOI: 10.1093/bioinformatics/btg362
Google Scholar
24
ModLoop: automated modeling of loops in protein structures
Fiser, Andras; Sali, Andrej
Bioinformatics (2003), 19 (18), 2500-2501CODEN: BOINFP; ISSN:1367-4803. (Oxford University Press)
Summary: ModLoop is a web server for automated modeling of loops in protein structures. The input is the at. coordinates of the protein structure in the Protein Data Bank format, and the specification of the starting and ending residues of one or more segments to be modeled, contg. no more than 20 residues in total. The output is the coordinates of the non-hydrogen atoms in the modeled segments. A user provides the input to the server via a simple web interface, and receives the output by e-mail. The server relies on the loop modeling routine in MODELLER that predicts the loop conformations by satisfaction of spatial restraints, without relying on a database of known protein structures. For a rapid response, ModLoop runs on a cluster of Linux PC computers.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXpvVSisLk%253D&md5=4b0fdaca0a412715682496883ab2019c
25
Sohl, C. D.; Ryan, M. R.; Luo, B.; Frey, K. M.; Anderson, K. S. Illuminating the molecular mechanisms of tyrosine kinase inhibitor resistance for the FGFR1 gatekeeper mutation: the Achilles’ heel of targeted therapy. ACS Chem. Biol. 2015, 10, 1319– 1329, DOI: 10.1021/acschembio.5b00014
Google Scholar
25
Illuminating the Molecular Mechanisms of Tyrosine Kinase Inhibitor Resistance for the FGFR1 Gatekeeper Mutation: The Achilles' Heel of Targeted Therapy
Sohl, Christal D.; Ryan, Molly R.; Luo, BeiBei; Frey, Kathleen M.; Anderson, Karen S.
ACS Chemical Biology (2015), 10 (5), 1319-1329CODEN: ACBCCT; ISSN:1554-8929. (American Chemical Society)
Human fibroblast growth factor receptors (FGFRs) 1-4 are a family of receptor tyrosine kinases that can serve as drivers of tumorigenesis. In particular, FGFR1 gene amplification has been implicated in squamous cell lung and breast cancers. Tyrosine kinase inhibitors (TKIs) targeting FGFR1, including AZD4547 and E3810 (Lucitanib), are currently in early phase clin. trials. Unfortunately, drug resistance limits the long-term success of TKIs, with mutations at the "gatekeeper" residue leading to tumor progression. Here we show the first structural and kinetic characterization of the FGFR1 gatekeeper mutation, V561M FGFR1. The V561M mutation confers a 38-fold increase in autophosphorylation achieved at least in part by a network of interacting residues forming a hydrophobic spine to stabilize the active conformation. Moreover, kinetic assays established that the V561M mutation confers significant resistance to E3810, while retaining affinity for AZD4547. Structural analyses of these TKIs with wild type (WT) and gatekeeper mutant forms of FGFR1 offer clues to developing inhibitors that maintain potency against gatekeeper mutations. We show that AZD4547 affinity is preserved by V561M FGFR1 due to a flexible linker that allows multiple inhibitor binding modes. This is the first example of a TKI binding in distinct conformations to WT and gatekeeper mutant forms of FGFR, highlighting adaptable regions in both the inhibitor and binding pocket crucial for drug design. Exploiting inhibitor flexibility to overcome drug resistance has been a successful strategy for combating diseases such as AIDS and may be an important approach for designing inhibitors effective against kinase gatekeeper mutations.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXislGitLo%253D&md5=6c13f193e9e4c3f38b1e59f46fbc6342
26
Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03; Gaussian, Inc.: Wallingford, CT, 2004.
Google Scholar
There is no corresponding record for this reference.
27
Case, D. A.; Cheatham, T. E.; Darden, T.; Gohlke, H.; Luo, R.; Merz, K. M.; Onufriev, A.; Simmerling, C.; Wang, B.; Woods, R. J. The Amber biomolecular simulation programs. J. Comput. Chem. 2005, 26, 1668– 1688, DOI: 10.1002/jcc.20290
Google Scholar
27
The amber biomolecular simulation programs
Case, David A.; Cheatham, Thomas E., III; Darden, Tom; Gohlke, Holger; Luo, Ray; Merz, Kenneth M., Jr.; Onufriev, Alexey; Simmerling, Carlos; Wang, Bing; Woods, Robert J.
Journal of Computational Chemistry (2005), 26 (16), 1668-1688CODEN: JCCHDD; ISSN:0192-8651. (John Wiley & Sons, Inc.)
The authors describe the development, current features, and some directions for future development of the Amber package of computer programs. This package evolved from a program that was constructed in the late 1970s to do Assisted Model Building with Energy Refinement, and now contains a group of programs embodying a no. of powerful tools of modern computational chem., focused on mol. dynamics and free energy calcns. of proteins, nucleic acids, and carbohydrates.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2MXht1SlsbbM&md5=93be29ff894bab96c783d24e9886c7d0
28
Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. Development and testing of a general amber force field. J. Comput. Chem. 2004, 25, 1157– 1174, DOI: 10.1002/jcc.20035
Google Scholar
28
Development and testing of a general Amber force field
Wang, Junmei; Wolf, Romain M.; Caldwell, James W.; Kollman, Peter A.; Case, David A.
Journal of Computational Chemistry (2004), 25 (9), 1157-1174CODEN: JCCHDD; ISSN:0192-8651. (John Wiley & Sons, Inc.)
We describe here a general Amber force field (GAFF) for org. mols. GAFF is designed to be compatible with existing Amber force fields for proteins and nucleic acids, and has parameters for most org. and pharmaceutical mols. that are composed of H, C, N, O, S, P, and halogens. It uses a simple functional form and a limited no. of atom types, but incorporates both empirical and heuristic models to est. force consts. and partial at. charges. The performance of GAFF in test cases is encouraging. In test I, 74 crystallog. structures were compared to GAFF minimized structures, with a root-mean-square displacement of 0.26 Å, which is comparable to that of the Tripos 5.2 force field (0.25 Å) and better than those of MMFF 94 and CHARMm (0.47 and 0.44 Å, resp.). In test II, gas phase minimizations were performed on 22 nucleic acid base pairs, and the minimized structures and intermol. energies were compared to MP2/6-31G* results. The RMS of displacements and relative energies were 0.25 Å and 1.2 kcal/mol, resp. These data are comparable to results from Parm99/RESP (0.16 Å and 1.18 kcal/mol, resp.), which were parameterized to these base pairs. Test III looked at the relative energies of 71 conformational pairs that were used in development of the Parm99 force field. The RMS error in relative energies (compared to expt.) is about 0.5 kcal/mol. GAFF can be applied to wide range of mols. in an automatic fashion, making it suitable for rational drug design and database searching.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2cXksFakurc%253D&md5=2992017a8cf51f89290ae2562403b115
29
Maier, J. A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.; Hauser, K. E.; Simmerling, C. ff14SB: Improving the accuracy of protein side chain and backbone parameters from ff99SB. J. Chem. Theory Comput. 2015, 11, 3696– 3713, DOI: 10.1021/acs.jctc.5b00255
Google Scholar
29
ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB
Maier, James A.; Martinez, Carmenza; Kasavajhala, Koushik; Wickstrom, Lauren; Hauser, Kevin E.; Simmerling, Carlos
Journal of Chemical Theory and Computation (2015), 11 (8), 3696-3713CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
Mol. mechanics is powerful for its speed in atomistic simulations, but an accurate force field is required. The Amber ff99SB force field improved protein secondary structure balance and dynamics from earlier force fields like ff99, but weaknesses in side chain rotamer and backbone secondary structure preferences have been identified. Here, we performed a complete refit of all amino acid side chain dihedral parameters, which had been carried over from ff94. The training set of conformations included multidimensional dihedral scans designed to improve transferability of the parameters. Improvement in all amino acids was obtained as compared to ff99SB. Parameters were also generated for alternate protonation states of ionizable side chains. Av. errors in relative energies of pairs of conformations were under 1.0 kcal/mol as compared to QM, reduced 35% from ff99SB. We also took the opportunity to make empirical adjustments to the protein backbone dihedral parameters as compared to ff99SB. Multiple small adjustments of φ and ψ parameters were tested against NMR scalar coupling data and secondary structure content for short peptides. The best results were obtained from a phys. motivated adjustment to the φ rotational profile that compensates for lack of ff99SB QM training data in the β-ppII transition region. Together, these backbone and side chain modifications (hereafter called ff14SB) not only better reproduced their benchmarks, but also improved secondary structure content in small peptides and reprodn. of NMR χ1 scalar coupling measurements for proteins in soln. We also discuss the Amber ff12SB parameter set, a preliminary version of ff14SB that includes most of its improvements.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhtFequ7rN&md5=7b803577b3b6912cc6750cfbd356596e
30
Jo, S.; Jiang, W. A generic implementation of replica exchange with solute tempering (REST2) algorithm in NAMD for complex biophysical simulations. Comput. Phys. Commun. 2015, 197, 304– 311, DOI: 10.1016/j.cpc.2015.08.030
Google Scholar
30
A generic implementation of replica exchange with solute tempering (REST2) algorithm in NAMD for complex biophysical simulations
Jo, Sunhwan; Jiang, Wei
Computer Physics Communications (2015), 197 (), 304-311CODEN: CPHCBZ; ISSN:0010-4655. (Elsevier B.V.)
Replica Exchange with Solute Tempering (REST2) is a powerful sampling enhancement algorithm of mol. dynamics (MD) in that it needs significantly smaller no. of replicas but achieves higher sampling efficiency relative to std. temp. exchange algorithm. In this paper, we extend the applicability of REST2 for quant. biophys. simulations through a robust and generic implementation in greatly scalable MD software NAMD. The rescaling procedure of force field parameters controlling REST2 "hot region" is implemented into NAMD at the source code level. A user can conveniently select hot region through VMD and write the selection information into a PDB file. The rescaling keyword/parameter is written in NAMD Tcl script interface that enables an on-the-fly simulation parameter change. Our implementation of REST2 is within communication-enabled Tcl script built on top of Charm++, thus communication overhead of an exchange attempt is vanishingly small. Such a generic implementation facilitates seamless cooperation between REST2 and other modules of NAMD to provide enhanced sampling for complex biomol. simulations. Three challenging applications including native REST2 simulation for peptide folding-unfolding transition, free energy perturbation/REST2 for abs. binding affinity of protein-ligand complex and umbrella sampling/REST2 Hamiltonian exchange for free energy landscape calcn. were carried out on IBM Blue Gene/Q supercomputer to demonstrate efficacy of REST2 based on the present implementation.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhsFSlu73I&md5=776ea438c536bf84540b1443494e647c
31
Ballester, P. J.; Mitchell, J. B. O. A machine learning approach to predicting protein–ligand binding affinity with applications to molecular docking. Bioinformatics 2010, 26, 1169– 1175, DOI: 10.1093/bioinformatics/btq112
Google Scholar
31
A machine learning approach to predicting protein-ligand binding affinity with applications to molecular docking
Ballester, Pedro J.; Mitchell, John B. O.
Bioinformatics (2010), 26 (9), 1169-1175CODEN: BOINFP; ISSN:1367-4803. (Oxford University Press)
Accurately predicting the binding affinities of large sets of diverse protein-ligand complexes is an extremely challenging task. The scoring functions that attempt such computational prediction are essential for analyzing the outputs of mol. docking, which in turn is an important technique for drug discovery, chem. biol. and structural biol. Each scoring function assumes a predetd. theory-inspired functional form for the relationship between the variables that characterize the complex, which also include parameters fitted to exptl. or simulation data and its predicted binding affinity. The inherent problem of this rigid approach is that it leads to poor predictivity for those complexes that do not conform to the modeling assumptions. Moreover, resampling strategies, such as cross-validation or bootstrapping, are still not systematically used to guard against the overfitting of calibration data in parameter estn. for scoring functions. We propose a novel scoring function (RF-Score) that circumvents the need for problematic modeling assumptions via non-parametric machine learning. In particular, Random Forest was used to implicitly capture binding effects that are hard to model explicitly. RF-Score is compared with the state of the art on the demanding PDBbind benchmark. Results show that RF-Score is a very competitive scoring function. Importantly, RF-Score's performance was shown to improve dramatically with training set size and hence the future availability of more high-quality structural and interaction data is expected to lead to improved versions of RF-Score.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3cXlt1Cjs78%253D&md5=6d2d48344dc5ba70f0ff85f432d0ec63
32
Jimenez, J.; Skalic, M.; Martinez-Rosell, G.; De Fabritiis, G. KDEEP: Protein–Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks. J. Chem. Inf. Model. 2018, 58, 287– 296, DOI: 10.1021/acs.jcim.7b00650
Google Scholar
32
KDEEP: Protein-Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks
Jimenez, Jose; Skalic, Miha; Martinez-Rosell, Gerard; De Fabritiis, Gianni
Journal of Chemical Information and Modeling (2018), 58 (2), 287-296CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
Accurately predicting protein-ligand binding affinities is an important problem in computational chem. since it can substantially accelerate drug discovery for virtual screening and lead optimization. We propose here a fast machine-learning approach for predicting binding affinities using state-of-the-art 3D-convolutional neural networks and compare this approach to other machine-learning and scoring methods using several diverse data sets. The results for the std. PDBbind (v.2016) core test-set are state-of-the-art with a Pearson's correlation coeff. of 0.82 and a RMSE of 1.27 in pK units between exptl. and predicted affinity, but accuracy is still very sensitive to the specific protein used. KDEEP is made available via PlayMol.org for users to test easily their own protein-ligand complexes, with each prediction taking a fraction of a second. We believe that the speed, performance, and ease of use of KDEEP makes it already an attractive scoring function for modern computational chem. pipelines.
>> More from SciFinder ^®
https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXkslSitQ%253D%253D&md5=81943a6732be99e5439e1e3a25fa4414
33
Coveney, P. V.; Dougherty, E. R.; Highfield, R. R. Big data need big theory too. Philos. Trans. R. Soc., A 2016, 374, 20160153, DOI: 10.1098/rsta.2016.0153
Google Scholar
There is no corresponding record for this reference.

Cited By

ARTICLE SECTIONS

Jump To

This article is cited by 27 publications.

Shunzhou Wan, Agastya P. Bhati, Peter V. Coveney. Comparison of Equilibrium and Nonequilibrium Approaches for Relative Binding Free Energy Predictions. Journal of Chemical Theory and Computation 2023, 19 (21) , 7846-7860. https://doi.org/10.1021/acs.jctc.3c00842
Christopher Lockhart, Xingyu Luo, Audrey Olson, Bryan M. Delfing, Xavier E. Laracuente, Kenneth W. Foreman, Mikell Paige, Kylene Kehn-Hall, Dmitri K. Klimov. Can Free Energy Perturbation Simulations Coupled with Replica-Exchange Molecular Dynamics Study Ligands with Distributed Binding Sites?. Journal of Chemical Information and Modeling 2023, 63 (15) , 4791-4802. https://doi.org/10.1021/acs.jcim.3c00631
Wei Jiang. Enhanced Configurational Sampling Approaches to Alchemical Ligand Binding Free Energy Simulations: Current Status and Challenges. The Journal of Physical Chemistry B 2023, 127 (31) , 6835-6841. https://doi.org/10.1021/acs.jpcb.3c02020
Agastya P. Bhati, Art Hoti, Andrew Potterton, Mateusz K. Bieniek, Peter V. Coveney. Long Time Scale Ensemble Methods in Molecular Dynamics: Ligand–Protein Interactions and Allostery in SARS-CoV-2 Targets. Journal of Chemical Theory and Computation 2023, 19 (11) , 3359-3378. https://doi.org/10.1021/acs.jctc.3c00020
Tai-Sung Lee, Hsu-Chun Tsai, Abir Ganguly, Darrin M. York. ACES: Optimized Alchemically Enhanced Sampling. Journal of Chemical Theory and Computation 2023, 19 (2) , 472-487. https://doi.org/10.1021/acs.jctc.2c00697
Alexander D. Wade, Agastya P. Bhati, Shunzhou Wan, Peter V. Coveney. Alchemical Free Energy Estimators and Molecular Dynamics Engines: Accuracy, Precision, and Reproducibility. Journal of Chemical Theory and Computation 2022, 18 (6) , 3972-3987. https://doi.org/10.1021/acs.jctc.2c00114
Agastya P. Bhati, Peter V. Coveney. Large Scale Study of Ligand–Protein Relative Binding Free Energy Calculations: Actionable Predictions from Statistically Robust Protocols. Journal of Chemical Theory and Computation 2022, 18 (4) , 2687-2702. https://doi.org/10.1021/acs.jctc.1c01288
Mateusz K. Bieniek, Agastya P. Bhati, Shunzhou Wan, Peter V. Coveney. TIES 20: Relative Binding Free Energy with a Flexible Superimposition Algorithm and Partial Ring Morphing. Journal of Chemical Theory and Computation 2021, 17 (2) , 1250-1265. https://doi.org/10.1021/acs.jctc.0c01179
Yaozong Li, Kwangho Nam. Repulsive Soft-Core Potentials for Efficient Alchemical Free Energy Calculations. Journal of Chemical Theory and Computation 2020, 16 (8) , 4776-4789. https://doi.org/10.1021/acs.jctc.0c00163
Rhys Evans, Ladislav Hovan, Gareth A. Tribello, Benjamin P. Cossins, Carolina Estarellas, Francesco L. Gervasio. Combining Machine Learning and Enhanced Sampling Techniques for Efficient and Accurate Calculation of Absolute Binding Free Energies. Journal of Chemical Theory and Computation 2020, 16 (7) , 4641-4654. https://doi.org/10.1021/acs.jctc.0c00075
Kyriakos Georgiou, Athina Konstantinidi, Johanna Hutterer, Kathrin Freudenberger, Felix Kolarov, George Lambrinidis, Ioannis Stylianakis, Margarita Stampelou, Günter Gauglitz, Antonios Kolocouris. Accurate calculation of affinity changes to the close state of influenza A M2 transmembrane domain in response to subtle structural changes of adamantyl amines using free energy perturbation methods in different lipid bilayers. Biochimica et Biophysica Acta (BBA) - Biomembranes 2024, 1866 (2) , 184258. https://doi.org/10.1016/j.bbamem.2023.184258
Martiniano Bello, Cindy Bandala. Evaluating the ability of end-point methods to predict the binding affinity tendency of protein kinase inhibitors. RSC Advances 2023, 13 (36) , 25118-25128. https://doi.org/10.1039/D3RA04916G
Yang Yu, Zhe Wang, Lingling Wang, Sheng Tian, Tingjun Hou, Huiyong Sun. Predicting the mutation effects of protein–ligand interactions via end-point binding free energy calculations: strategies and analyses. Journal of Cheminformatics 2022, 14 (1) https://doi.org/10.1186/s13321-022-00639-y
Alice E. Brankin, Philip W. Fowler. Predicting antibiotic resistance in complex protein targets using alchemical free energy methods. Journal of Computational Chemistry 2022, 43 (26) , 1771-1782. https://doi.org/10.1002/jcc.26979
Emilia P. Barros, Benjamin Ries, Lennard Böselt, Candide Champion, Sereina Riniker. Recent developments in multiscale free energy simulations. Current Opinion in Structural Biology 2022, 72 , 55-62. https://doi.org/10.1016/j.sbi.2021.08.003
Ran Friedman. Computational studies of protein–drug binding affinity changes upon mutations in the drug target. WIREs Computational Molecular Science 2022, 12 (1) https://doi.org/10.1002/wcms.1563
Agastya P. Bhati, Shunzhou Wan, Dario Alfè, Austin R. Clyde, Mathis Bode, Li Tan, Mikhail Titov, Andre Merzky, Matteo Turilli, Shantenu Jha, Roger R. Highfield, Walter Rocchia, Nicola Scafuri, Sauro Succi, Dieter Kranzlmüller, Gerald Mathias, David Wifling, Yann Donon, Alberto Di Meglio, Sofia Vallecorsa, Heng Ma, Anda Trifan, Arvind Ramanathan, Tom Brettin, Alexander Partin, Fangfang Xia, Xiaotan Duan, Rick Stevens, Peter V. Coveney. Pandemic drugs at pandemic speed: infrastructure for accelerating COVID-19 drug discovery with hybrid machine learning- and physics-based simulations on high-performance computers. Interface Focus 2021, 11 (6) https://doi.org/10.1098/rsfs.2021.0018
Shunzhou Wan, Deepak Kumar, Valentin Ilyin, Ussama Al Homsi, Gulab Sher, Alexander Knuth, Peter V. Coveney. The effect of protein mutations on drug binding suggests ensuing personalised drug selection. Scientific Reports 2021, 11 (1) https://doi.org/10.1038/s41598-021-92785-w
Aparna Vilas Dongre, Sudip Das, Asutosh Bellur, Sanjeev Kumar, Anusha Chandrashekarmath, Tarak Karmakar, Padmanabhan Balaram, Sundaram Balasubramanian, Hemalatha Balaram. Structural basis for the hyperthermostability of an archaeal enzyme induced by succinimide formation. Biophysical Journal 2021, 120 (17) , 3732-3746. https://doi.org/10.1016/j.bpj.2021.07.014
Edward King, Erick Aitchison, Han Li, Ray Luo. Recent Developments in Free Energy Calculations for Drug Discovery. Frontiers in Molecular Biosciences 2021, 8 https://doi.org/10.3389/fmolb.2021.712085
Himanshu Goel, Anthony Hazel, Vincent D. Ustach, Sunhwan Jo, Wenbo Yu, Alexander D. MacKerell. Rapid and accurate estimation of protein–ligand relative binding affinities using site-identification by ligand competitive saturation. Chemical Science 2021, 12 (25) , 8844-8858. https://doi.org/10.1039/D1SC01781K
Shunzhou Wan, Robert C. Sinclair, Peter V. Coveney. Uncertainty quantification in classical molecular dynamics. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 2021, 379 (2197) https://doi.org/10.1098/rsta.2020.0082
Natacha Gillet, Alessio Bartocci, Elise Dumont. Assessing the sequence dependence of pyrimidine–pyrimidone (6–4) photoproduct in a duplex double-stranded DNA: A pitfall for microsecond range simulation. The Journal of Chemical Physics 2021, 154 (13) https://doi.org/10.1063/5.0041332
Shunzhou Wan, Andrew Potterton, Fouad S. Husseini, David W. Wright, Alexander Heifetz, Maciej Malawski, Andrea Townsend-Nicholson, Peter V. Coveney. Hit-to-lead and lead optimization binding free energy calculations for G protein-coupled receptors. Interface Focus 2020, 10 (6) , 20190128. https://doi.org/10.1098/rsfs.2019.0128
Shunzhou Wan, Agastya P. Bhati, Stefan J. Zasada, Peter V. Coveney. Rapid, accurate, precise and reproducible ligand–protein binding free energy prediction. Interface Focus 2020, 10 (6) , 20200007. https://doi.org/10.1098/rsfs.2020.0007
Robert Hall, Tom Dixon, Alex Dickson. On Calculating Free Energy Differences Using Ensembles of Transition Paths. Frontiers in Molecular Biosciences 2020, 7 https://doi.org/10.3389/fmolb.2020.00106
Shunzhou Wan, Gary Tresadern, Laura Pérez‐Benito, Herman van Vlijmen, Peter V. Coveney. Accuracy and Precision of Alchemical Relative Free Energy Predictions with and without Replica‐Exchange. Advanced Theory and Simulations 2020, 3 (1) https://doi.org/10.1002/adts.201900195

Download PDF

Abstract
High Resolution Image
Download MS PowerPoint Slide
Figure 1
Figure 1. Structures of FGFR3 and inhibitors studied in this work: (a) the binding site of tyrosine kinase domain for FGFR3 in complex with ACP, an ATP-analogue (PDB ID: 4K33). FGFR3 is depicted in cartoon and ACP in bond representation. Mutations of three residues, V555, I538, and N540 (ball-and-stick representation), are among the most common genetic variants in FGFR3. The chemical structures of four ATP competitive inhibitors are shown in panels b–e: (b) AZD4547, (c) BGJ-398, (d) TKI258, and (e) JNJ4275649.
High Resolution Image
Download MS PowerPoint Slide
Figure 2
Figure 2. Different regions in the λ-REST2 simulations. The AZD4547-V555M complex is shown here as an example. The hybrid residue, denoted as the alchemical region, is depicted as a ball-and-stick model. It consists of disappearing (red) and appearing (blue) groups which are slightly separated for reasons of clarity. They can fully or partially overlap in the simulation as there are no interactions between them. The REST2 region, including the alchemical region (red and blue ball-and-stick), part of the ligand (orange bond), and surrounding protein residues (orange stick), is designated as the “hot” region. The selection of the REST2 region is described in the main text (section 2.3 REST2 region).
High Resolution Image
Download MS PowerPoint Slide
Figure 3
Figure 3. Comparison of the predicted ΔΔG_calc values using TIES (black circles), TIES-λ-REST2 (λR2, up/down triangles) with those from experiments for V555M mutant complexes with the highest T_eff of the chosen REST2 region at 600 K for receptor and complexes except those with AZD4547 which are at 1500 K (red triangles pointing up), and at 1500 K for receptor and 3000 K for complexes (blue triangles pointing down). Results of AZD4547 from the bent conformation are represented using filled circles and triangles. The dotted lines (x = 0 and y = 0) create four quadrants. Data points in quadrants I (x > 0 and y > 0) and III (x < 0 and y < 0) indicate that the calculated binding free energy differences agree directionally with those from the experimental data. The results from TIES-λ-REST2-M (λR2-M) are very close to those from λR2 (Table 1), and are not shown for reasons of clarity.
High Resolution Image
Download MS PowerPoint Slide
Figure 4
Figure 4. Comparison of the predicted ΔΔG_calc values using TIES (black circles), TIES-λ-REST2 (λR2, up/down triangles) with those from experiment for for all inhibitors bound to FGFR3: (a) I538V mutant and (b) N540S mutant, when the highest T_eff for the chosen REST2 region is at 800 K for receptor and 1500 K for complexes (red triangles pointing up) and at 1500 K for receptor and 3000 K for complexes (blue triangles pointing down). The dotted lines (x = 0 and y = 0) create four quadrants. Data points in quadrants I (x > 0 and y > 0) and III (x < 0 and y < 0) indicate that the ΔΔG_calc values agree directionally with ΔΔG_exp. The results from TIES-λ-REST2-M (λR2-M) are very close to those from λR2 (Table 1), and are not shown for reasons of clarity.
High Resolution Image
Download MS PowerPoint Slide
Figure 5
Figure 5. Variation of the cumulative average of ΔΔG_calc^TIES with simulation length for all inhibitors bound to the FGFR3 I538V mutant. The corresponding experimental value for each inhibitor is shown by a dashed line of the same color.
High Resolution Image
Download MS PowerPoint Slide
Figure 6
Figure 6. Variation of the cumulative average of ΔΔG calculated using schemes λR2 and λR2-M with simulation length for all complexes. The highest T_eff for receptor and complex are 1500 and 3000 K, respectively, in the case of I538V and N540S mutants, while the corresponding values in the case of V555M mutant are 600 and 600 K/1500 K. The corresponding experimental values for each inhibitor are shown by a dashed line of the same color.
High Resolution Image
Download MS PowerPoint Slide
Figure 7
Figure 7. Two distinct conformations of inhibitor AZD4547 found experimentally when bound to the FGFR gatekeeper mutant. The three hydrogen bonds, marked with black dashed lines and labeled as H1, H2, and H3, keep the middle portion of the inhibitor stable. The value of the dihedral angle between the four carbon atoms highlighted in orange can be used as an indicator of the occurrence of the two conformations. The atoms displayed as balls lie in the REST2 region while the ones displayed as lines reside outside it.
High Resolution Image
Download MS PowerPoint Slide
Figure 8
Figure 8. Normalized frequency distributions of the three hydrogen bonds (H1, H2, and H3 from Figure 7) in λ-REST2 trajectories at the λ = 1 end-point of the V555M–AZD4547 complexes when using the linear as well as the bent starting structures.
High Resolution Image
Download MS PowerPoint Slide
Figure 9
Figure 9. Normalized frequency distributions of the dihedral angle between the four carbon atoms highlighted in orange in Figure 7 for different λ states of V555M–AZD4547 complexes in standard TIES (in blue) as well as λ-REST2 simulations showing the relative populations of the two conformations of AZD4547. In the case of λ-REST2 simulations, the distributions from the first (1–4 ns; in black) and the last 4 ns (17–20 ns; in red) are shown separately.
High Resolution Image
Download MS PowerPoint Slide
Figure 10
Figure 10. Inhibitor TKI258 bound to V555M mutant. It forms two hydrogen bonds with the hinge region of the protein which are displayed with black dashed lines and labeled as H1 and H2. The atoms shown as balls lie in the “hot” region. The atoms are shown in the standard color code: carbon in green, oxygen in red, nitrogen in blue, hydrogen in white, and fluorine in pink.
High Resolution Image
Download MS PowerPoint Slide
Figure 11
Figure 11. Normalized frequency distributions of H1 and H2 from Figure 10 for the two end-points in the case of the standard TIES as well as λ-REST2 simulations of the V555M–TKI258 complex. λ-REST2 simulations sample a larger comformational space than TIES, as evidenced by the lower and wider distributions of the distances, and the second peaks in the λ = 1 end-point (V555M).
High Resolution Image
Download MS PowerPoint Slide
Figure 12
Figure 12. Normalized frequency distributions of H1 and H2 from Figure 10 for the two end-points in the case of standard TIES as well as λ-REST2 simulations of the I538V–TKI258 complex. The long tails and additional peaks beyond 4 Å indicate that λ-REST2 simulations sample some conformations irrelevant to stable inhibitor binding.
High Resolution Image
Download MS PowerPoint Slide
Figure 13
Figure 13. Normalized frequency distributions of H1 and H2 from Figure 10 for the two end-points in the case of standard TIES as well as λ-REST2 simulations of the N540S–TKI258 complex. The long tails and additional peaks beyond 4 Å indicate that λ-REST2 simulations sample some conformations irrelevant to stable inhibitor binding.
High Resolution Image
Download MS PowerPoint Slide
References
ARTICLE SECTIONS
Jump To
This article references 33 other publications.
1. 1
  Wright, D. W.; Wan, S.; Shublaq, N.; Zasada, S. J.; Coveney, P. V. From base pair to bedside: molecular simulation and the translation of genomics to personalized medicine. WIREs Syst. Biol. Med. 2012, 4, 585– 598, DOI: 10.1002/wsbm.1186
  Google Scholar
  1
  From base pair to bedside: molecular simulation and the translation of genomics to personalized medicine
  Wright, David W.; Wan, Shunzhou; Shublaq, Nour; Zasada, Stefan J.; Coveney, Peter V.
  Wiley Interdisciplinary Reviews: Systems Biology and Medicine (2012), 4 (6), 585-598CODEN: WIRSBW; ISSN:1939-005X. (Wiley-Blackwell)
  A review. Despite the promises made that genomic sequencing would transform therapy by introducing a new era of personalized medicine, relatively few tangible breakthroughs have been made. This has led to the recognition that complex interactions at multiple spatial, temporal, and organizational levels may often combine to produce disease. Understanding this complexity requires that existing and future models are used and interpreted within a framework that incorporates knowledge derived from investigations at multiple levels of biol. function. It also requires a computational infrastructure capable of dealing with the vast quantities of data generated by genomic approaches. In this review, we discuss the use of mol. modeling to generate quant. and qual. insights at the smallest scales of the systems biol. hierarchy, how it can play an important role in the development of a systems understanding of disease and in the application of such knowledge to help discover new therapies and target existing ones on a personal level.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3sXpslSgsg%253D%253D&md5=b6bf5cb6b9fab33543986b9f8ceed6ec
2. 2
  Wang, L. Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field. J. Am. Chem. Soc. 2015, 137, 2695– 2703, DOI: 10.1021/ja512751q
  Google Scholar
  2
  Accurate and Reliable Prediction of Relative Ligand Binding Potency in Prospective Drug Discovery by Way of a Modern Free-Energy Calculation Protocol and Force Field
  Wang, Lingle; Wu, Yujie; Deng, Yuqing; Kim, Byungchan; Pierce, Levi; Krilov, Goran; Lupyan, Dmitry; Robinson, Shaughnessy; Dahlgren, Markus K.; Greenwood, Jeremy; Romero, Donna L.; Masse, Craig; Knight, Jennifer L.; Steinbrecher, Thomas; Beuming, Thijs; Damm, Wolfgang; Harder, Ed; Sherman, Woody; Brewer, Mark; Wester, Ron; Murcko, Mark; Frye, Leah; Farid, Ramy; Lin, Teng; Mobley, David L.; Jorgensen, William L.; Berne, Bruce J.; Friesner, Richard A.; Abel, Robert
  Journal of the American Chemical Society (2015), 137 (7), 2695-2703CODEN: JACSAT; ISSN:0002-7863. (American Chemical Society)
  Designing tight-binding ligands is a primary objective of small-mol. drug discovery. Over the past few decades, free-energy calcns. have benefited from improved force fields and sampling algorithms, as well as the advent of low-cost parallel computing. However, it has proven to be challenging to reliably achieve the level of accuracy that would be needed to guide lead optimization (∼5× in binding affinity) for a wide range of ligands and protein targets. Not surprisingly, widespread com. application of free-energy simulations has been limited due to the lack of large-scale validation coupled with the tech. challenges traditionally assocd. with running these types of calcns. Here, we report an approach that achieves an unprecedented level of accuracy across a broad range of target classes and ligands, with retrospective results encompassing 200 ligands and a wide variety of chem. perturbations, many of which involve significant changes in ligand chem. structures. In addn., we have applied the method in prospective drug discovery projects and found a significant improvement in the quality of the compds. synthesized that have been predicted to be potent. Compds. predicted to be potent by this approach have a substantial redn. in false positives relative to compds. synthesized on the basis of other computational or medicinal chem. approaches. Furthermore, the results are consistent with those obtained from our retrospective studies, demonstrating the robustness and broad range of applicability of this approach, which can be used to drive decisions in lead optimization.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhsF2iuro%253D&md5=37a4f4a6c085f47ed531342643b6c33b
3. 3
  Bernardi, R. C.; Melo, M. C.; Schulten, K. Enhanced sampling techniques in molecular dynamics simulations of biological systems. Biochim. Biophys. Acta, Gen. Subj. 2015, 1850, 872– 877, DOI: 10.1016/j.bbagen.2014.10.019
  Google Scholar
  3
  Enhanced sampling techniques in molecular dynamics simulations of biological systems
  Bernardi, Rafael C.; Melo, Marcelo C. R.; Schulten, Klaus
  Biochimica et Biophysica Acta, General Subjects (2015), 1850 (5), 872-877CODEN: BBGSB3; ISSN:0304-4165. (Elsevier B.V.)
  A review. Mol. dynamics has emerged as an important research methodol. covering systems to the level of millions of atoms. However, insufficient sampling often limits its application. The limitation is due to rough energy landscapes, with many local min. sepd. by high-energy barriers, which govern the biomol. motion. In the past few decades methods have been developed that address the sampling problem, such as replica-exchange mol. dynamics, metadynamics and simulated annealing. Here the authors present an overview over theses sampling methods in an attempt to shed light on which should be selected depending on the type of system property studied. Enhanced sampling methods have been employed for a broad range of biol. systems and the choice of a suitable method is connected to biol. and phys. characteristics of the system, in particular system size. While metadynamics and replica-exchange mol. dynamics are the most adopted sampling methods to study biomol. dynamics, simulated annealing is well suited to characterize very flexible systems. The use of annealing methods for a long time was restricted to simulation of small proteins; however, a variant of the method, generalized simulated annealing, can be employed at a relatively low computational cost to large macromol. complexes. Mol. dynamics trajectories frequently do not reach all relevant conformational substates, for example those connected with biol. function, a problem that can be addressed by employing enhanced sampling algorithms. This article is part of a Special Issue entitled Recent developments of mol. dynamics.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2cXhvVanurbN&md5=44e3ac1aa042c35ee08aa86a6a63e78e
4. 4
  Valsson, O.; Parrinello, M. Variational approach to enhanced sampling and free energy calculations. Phys. Rev. Lett. 2014, 113, 090601, DOI: 10.1103/PhysRevLett.113.090601
  Google Scholar
  4
  Variational approach to enhanced sampling and free energy calculations
  Valsson, Omar; Parrinello, Michele
  Physical Review Letters (2014), 113 (9), 090601CODEN: PRLTAO; ISSN:0031-9007. (American Physical Society)
  The ability of widely used sampling methods, such as mol. dynamics or Monte Carlo simulations, to explore complex free energy landscapes is severely hampered by the presence of kinetic bottlenecks. A large no. of solns. have been proposed to alleviate this problem. Many are based on the introduction of a bias potential which is a function of a small no. of collective variables. However constructing such a bias is not simple. Here we introduce a functional of the bias potential and an assocd. variational principle. The bias that minimizes the functional relates in a simple way to the free energy surface. This variational principle can be turned into a practical, efficient, and flexible sampling method. A no. of numerical examples are presented which include the detn. of a three-dimensional free energy surface. We argue that, beside being numerically advantageous, our variational approach provides a convenient and novel standpoint for looking at the sampling problem.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2cXhslShtL7O&md5=4d66ba6f0d445e693e28dbeae09cd936
5. 5
  Fukunishi, H.; Watanabe, O.; Takada, S. On the Hamiltonian replica exchange method for efficient sampling of biomolecular systems: Application to protein structure prediction. J. Chem. Phys. 2002, 116, 9058– 9067, DOI: 10.1063/1.1472510
  Google Scholar
  5
  On the Hamiltonian replica exchange method for efficient sampling of biomolecular systems: Application to protein structure prediction
  Fukunishi, Hiroaki; Watanabe, Osamu; Takada, Shoji
  Journal of Chemical Physics (2002), 116 (20), 9058-9067CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
  Motivated by the protein structure prediction problem, we develop two variants of the Hamiltonian replica exchange methods (REMs) for efficient configuration sampling, (1) the scaled hydrophobicity REM and (2) the phantom chain REM, and compare their performance with the ordinary REM. We first point out that the ordinary REM has a shortage for the application to large systems such as biomols. and that the Hamiltonian REM, an alternative formulation of the REM, can give a remedy for it. We then propose two examples of the Hamiltonian REM that are suitable for a coarse-grained protein model. (1) The scaled hydrophobicity REM preps. replicas that are characterized by various strengths of hydrophobic interaction. The strongest interaction that mimics aq. soln. environment makes proteins folding, while weakened hydrophobicity unfolds proteins as in org. solvent. Exchange between these environments enables proteins to escape from misfolded traps and accelerate conformational search. This resembles the roles of mol. chaperone that assist proteins to fold in vivo. (2) The phantom chain REM uses replicas that allow various degrees of at. overlaps. By allowing at. overlap in some of replicas, the peptide chain can cross over itself, which can accelerate conformation sampling. Using a coarse-gained model we developed, we compute equil. probability distributions for poly-alanine 16-mer and for a small protein by these REMs and compare the accuracy of the results. We see that the scaled hydrophobicity REM is the most efficient method among the three REMs studied.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD38XjsFKmsLo%253D&md5=7ac571a5afdd63b0b4b29cfdec06f53b
6. 6
  Wang, L.; Friesner, R. A.; Berne, B. J. Replica exchange with solute scaling: A more efficient version of replica exchange with solute tempering (REST2). J. Phys. Chem. B 2011, 115, 9431– 9438, DOI: 10.1021/jp204407d
  Google Scholar
  6
  Replica Exchange with Solute Scaling: A More Efficient Version of Replica Exchange with Solute Tempering (REST2)
  Wang, Lingle; Friesner, Richard A.; Berne, B. J.
  Journal of Physical Chemistry B (2011), 115 (30), 9431-9438CODEN: JPCBFK; ISSN:1520-5207. (American Chemical Society)
  A small change in the Hamiltonian scaling in Replica Exchange with Solute Tempering (REST) is found to improve its sampling efficiency greatly, esp. for the sampling of aq. protein solns. in which there are large-scale solute conformation changes. Like the original REST (REST1), the new version (which the authors call REST2) also bypasses the poor scaling with system size of the std. Temp. Replica Exchange Method (TREM), reducing the no. of replicas (parallel processes) from what must be used in TREM. This redn. is accomplished by deforming the Hamiltonian function for each replica in such a way that the acceptance probability for the exchange of replica configurations does not depend on the no. of explicit water mols. in the system. For proof of concept, REST2 is compared with TREM and with REST1 for the folding of the trpcage and β-hairpin in water. The comparisons confirm that REST2 greatly reduces the no. of CPUs required by regular replica exchange and greatly increases the sampling efficiency over REST1. This method reduces the CPU time required for calcg. thermodn. avs. and for the ab initio folding of proteins in explicit water.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXosFymurs%253D&md5=8acb0d670bcb70eacfc30e32d5dfb6dd
7. 7
  Wang, L.; Berne, B. J.; Friesner, R. A. On achieving high accuracy and reliability in the calculation of relative protein–ligand binding affinities. Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 1937– 1942, DOI: 10.1073/pnas.1114017109
  Google Scholar
  7
  On achieving high accuracy and reliability in the calculation of relative protein-ligand binding affinities
  Wang, Lingle; Berne, B. J.; Friesner, Richard A.
  Proceedings of the National Academy of Sciences of the United States of America (2012), 109 (6), 1937-1942, S1937/1-S1937/7CODEN: PNASA6; ISSN:0027-8424. (National Academy of Sciences)
  We apply a free energy perturbation simulation method, free energy perturbation/replica exchange with solute tempering, to two modifications of protein-ligand complexes that lead to significant conformational changes, the first in the protein and the second in the ligand. The approach is shown to facilitate sampling in these challenging cases where high free energy barriers sep. the initial and final conformations and leads to superior convergence of the free energy as demonstrated both by consistency of the results (independence from the starting conformation) and agreement with exptl. binding affinity data. The second case, consisting of two neutral thrombin ligands that are taken from a recent medicinal chem. program for this interesting pharmaceutical target, is of particular significance in that it demonstrates that good results can be obtained for large, complex ligands, as opposed to relatively simple model systems. To achieve quant. agreement with expt. in the thrombin case, a next generation force field, Optimized Potentials for Liq. Simulations 2.0, is required, which provides superior charges and torsional parameters as compared to earlier alternatives.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC38XivVSht74%253D&md5=23db7d18e624d5e1214bbd6702619a87
8. 8
  Paliwal, H.; Shirts, M. R. A benchmark test set for alchemical free energy transformations and its use to quantify error in common free energy methods. J. Chem. Theory Comput. 2011, 7, 4115– 4134, DOI: 10.1021/ct2003995
  Google Scholar
  8
  A Benchmark Test Set for Alchemical Free Energy Transformations and Its Use to Quantify Error in Common Free Energy Methods
  Paliwal, Himanshu; Shirts, Michael R.
  Journal of Chemical Theory and Computation (2011), 7 (12), 4115-4134CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
  There is a significant need for improved tools to validate thermophys. quantities computed via mol. simulation. In this paper we present the initial version of a benchmark set of testing methods for calcg. free energies of mol. transformation in soln. This set is based on mol. changes common to many mol. design problems, such as insertion and deletion of at. sites and changing at. partial charges. We use this benchmark set to compare the statistical efficiency, reliability, and quality of uncertainty ests. for a no. of published free energy methods, including thermodn. integration, free energy perturbation, the Bennett acceptance ratio (BAR) and its multistate equiv. MBAR. We identify MBAR as the consistently best performing method, though other methods are frequently comparable in reliability and accuracy in many cases. We demonstrate that assumptions of Gaussian distributed errors in free energies are usually valid for most methods studied. We demonstrate that bootstrap error estn. is a robust and useful technique for estg. statistical variance for all free energy methods studied. This benchmark set is provided in a no. of different file formats with the hope of becoming a useful and general tool for method comparisons.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3MXhtlaqtrvM&md5=41383373300216165a7f1724ad535451
9. 9
  Shirts, M. R.; Chodera, J. D. Statistically optimal analysis of samples from multiple equilibrium states. J. Chem. Phys. 2008, 129, 124105, DOI: 10.1063/1.2978177
  Google Scholar
  9
  Statistically optimal analysis of samples from multiple equilibrium states
  Shirts, Michael R.; Chodera, John D.
  Journal of Chemical Physics (2008), 129 (12), 124105/1-124105/10CODEN: JCPSA6; ISSN:0021-9606. (American Institute of Physics)
  We present a new estimator for computing free energy differences and thermodn. expectations as well as their uncertainties from samples obtained from multiple equil. states via either simulation or expt. The estimator, which we call the multistate Bennett acceptance ratio estimator (MBAR) because it reduces to the Bennett acceptance ratio estimator (BAR) when only two states are considered, has significant advantages over multiple histogram reweighting methods for combining data from multiple states. It does not require the sampled energy range to be discretized to produce histograms, eliminating bias due to energy binning and significantly reducing the time complexity of computing a soln. to the estg. equations in many cases. Addnl., an est. of the statistical uncertainty is provided for all estd. quantities. In the large sample limit, MBAR is unbiased and has the lowest variance of any known estimator for making use of equil. data collected from multiple states. We illustrate this method by producing a highly precise est. of the potential of mean force for a DNA hairpin system, combining data from multiple optical tweezer measurements under const. force bias. (c) 2008 American Institute of Physics.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXht1WnsL7F&md5=479183e1f45fc58dd7c6e5ef1e73d45d
10. 10
  Bhati, A. P.; Wan, S.; Wright, D. W.; Coveney, P. V. Rapid, accurate, precise, and reliable relative free energy prediction using ensemble based thermodynamic integration. J. Chem. Theory Comput. 2017, 13, 210– 222, DOI: 10.1021/acs.jctc.6b00979
  Google Scholar
  10
  Rapid, Accurate, Precise, and Reliable Relative Free Energy Prediction Using Ensemble Based Thermodynamic Integration
  Bhati, Agastya P.; Wan, Shunzhou; Wright, David W.; Coveney, Peter V.
  Journal of Chemical Theory and Computation (2017), 13 (1), 210-222CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
  The accurate prediction of the binding affinities of ligands to proteins is a major goal in drug discovery and personalized medicine. The time taken to make such predictions is of similar importance to their accuracy, precision and reliability. In the last few years, an ensemble based mol. dynamics approach has been proposed that provides a route to reliable predictions of free energies based on the mol. mechanics Poisson-Boltzmann surface area method which meets the requirements of accuracy, precision and reliability. Here, we describe an equiv. methodol. based on thermodn. integration to substantially improve the accuracy, precision and reliability of calcd. relative binding free energies. We report the performance of the method when applied to a diverse set of protein targets and ligands. The results are in very good agreement with exptl. data (90% of calcns. agree to within 1 kcal/mol) while the method is reproducible by construction. Statistical uncertainties of the order of 0.5 kcal/mol or less are achieved. We present a systematic account of how the uncertainty in the predictions may be estd.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitVGntrrJ&md5=510b70188112a3578030e291ce19b127
11. 11
  Bhati, A. P.; Wan, S.; Hu, Y.; Sherborne, B.; Coveney, P. V. Uncertainty quantification in alchemical free energy methods. J. Chem. Theory Comput. 2018, 14, 2867– 2880, DOI: 10.1021/acs.jctc.7b01143
  Google Scholar
  11
  Uncertainty Quantification in Alchemical Free Energy Methods
  Bhati, Agastya P.; Wan, Shunzhou; Hu, Yuan; Sherborne, Brad; Coveney, Peter V.
  Journal of Chemical Theory and Computation (2018), 14 (6), 2867-2880CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
  Alchem. free energy methods have gained much importance recently from several reports of improved ligand-protein binding affinity predictions based on their implementation using mol. dynamics simulations. A large no. of variants of such methods implementing different accelerated sampling techniques and free energy estimators are available, each claimed to be better than the others in its own way. However, the key features of reproducibility and quantification of assocd. uncertainties in such methods have barely been discussed. Here, we apply a systematic protocol for uncertainty quantification to a no. of popular alchem. free energy methods, covering both abs. and relative free energy predictions. We show that a reliable measure of error estn. is provided by ensemble simulation-an ensemble of independent MD simulations-which applies irresp. of the free energy method. The need to use ensemble methods is fundamental and holds regardless of the duration of time of the mol. dynamics simulations performed.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXotFaqtr4%253D&md5=5f80e23df9c8aafb58914af06de24e1f
12. 12
  Wan, S.; Bhati, A. P.; Skerratt, S.; Omoto, K.; Shanmugasundaram, V.; Bagal, S. K.; Coveney, P. V. Evaluation and characterization of Trk kinase inhibitors for the treatment of pain: reliable binding affinity predictions from theory and computation. J. Chem. Inf. Model. 2017, 57, 897– 909, DOI: 10.1021/acs.jcim.6b00780
  Google Scholar
  12
  Evaluation and Characterization of Trk Kinase Inhibitors for the Treatment of Pain: Reliable Binding Affinity Predictions from Theory and Computation
  Wan, Shunzhou; Bhati, Agastya P.; Skerratt, Sarah; Omoto, Kiyoyuki; Shanmugasundaram, Veerabahu; Bagal, Sharan K.; Coveney, Peter V.
  Journal of Chemical Information and Modeling (2017), 57 (4), 897-909CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
  Optimization of ligand binding affinity to the target protein of interest is a primary objective in small-mol. drug discovery. Until now, the prediction of binding affinities by computational methods has not been widely applied in the drug discovery process, mainly due to its lack of accuracy and reproducibility, as well as the long turnaround times required to obtain results. Herein, the authors report on a collaborative study that compares tropomyosin receptor kinase A (TrkA) binding affinity predictions using two recently formulated fast computational approaches - namely ESMACS (Enhanced Sampling of Mol. dynamics with Approxn. of Continuum Solvent) and TIES (Thermodn. Integration with Enhanced Sampling) - to exptl. derived TrkA binding affinities for a set of Pfizer pan-Trk compds. ESMACS gives precise and reproducible results and is applicable to highly diverse sets of compds. It also provides detailed chem. insight into the nature of ligand-protein binding. TIES can predict and thus optimize more subtle changes in binding affinities between compds. of similar structure. Individual binding affinities were calcd. in a few hours, exhibiting good correlations with the exptl. data of 0.79 and 0.88 from ESMACS and TIES approaches resp. The speed, level of accuracy and precision of the calcns. are such that the affinity predictions can be used to rapidly explain the effects of compd. modifications on TrkA binding affinity. The methods could therefore be used as tools to guide lead optimization efforts across multiple prospective structurally-enabled programs in the drug discovery setting for a wide range of compds. and targets.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2sXksVyrurc%253D&md5=2e366f56c1ee8c2f2b2760b9098e2030
13. 13
  Wan, S.; Bhati, A. P.; Zasada, S. J.; Wall, I.; Green, D.; Bamborough, P.; Coveney, P. V. Rapid and reliable binding affinity prediction of bromodomain inhibitors: A computational study. J. Chem. Theory Comput. 2017, 13, 784– 795, DOI: 10.1021/acs.jctc.6b00794
  Google Scholar
  13
  Rapid and Reliable Binding Affinity Prediction of Bromodomain Inhibitors: A Computational Study
  Wan, Shunzhou; Bhati, Agastya P.; Zasada, Stefan J.; Wall, Ian; Green, Darren; Bamborough, Paul; Coveney, Peter V.
  Journal of Chemical Theory and Computation (2017), 13 (2), 784-795CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
  Binding free energies of bromodomain inhibitors are calcd. with recently formulated approaches, namely ESMACS (enhanced sampling of mol. dynamics with approxn. of continuum solvent) and TIES (thermodn. integration with enhanced sampling). A set of compds. is provided by GlaxoSmithKline, which represents a range of chem. functionality and binding affinities. The predicted binding free energies exhibit a good Spearman correlation of 0.78 with the exptl. data from the 3-trajectory ESMACS, and an excellent correlation of 0.92 from the TIES approach where applicable. Given access to suitable high end computing resources and a high degree of automation, the authors can compute individual binding affinities in a few hours with precisions no greater than 0.2 kcal/mol for TIES, and no larger than 0.34 kcal/mol and 1.71 kcal/mol for the 1- and 3-trajectory ESMACS approaches.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitFektLnF&md5=713436084662420482684fb50db0832e
14. 14
  Porta, R.; Borea, R.; Coelho, A.; Khan, S.; Araújo, A.; Reclusa, P.; Franchina, T.; Steen, N. V. D.; Dam, P. V.; Ferri, J.; Sirera, R.; Naing, A.; Hong, D.; Rolfo, C. FGFR a promising druggable target in cancer: Molecular biology and new drugs. Crit. Rev. Oncol. Hematol. 2017, 113, 256– 267, DOI: 10.1016/j.critrevonc.2017.02.018
  Google Scholar
  14
  FGFR a promising druggable target in cancer: Molecular biology and new drugs
  Porta Rut; Borea Roberto; Coelho Andreia; Reclusa Pablo; Van Dam Peter; Ferri Jose; Sirera Rafael; Khan Shahanavaj; Araujo Antonio; Franchina Tindara; Van Der Steen Nele; Naing Aung; Hong David; Rolfo Christian
  Critical reviews in oncology/hematology (2017), 113 (), 256-267 ISSN:.
  INTRODUCTION: The Fibroblast Growth Factor Receptor (FGFR) family consists of Tyrosine Kinase Receptors (TKR) involved in several biological functions. Recently, alterations of FGFR have been reported to be important for progression and development of several cancers. In this setting, different studies are trying to evaluate the efficacy of different therapies targeting FGFR. AREAS COVERED: This review summarizes the current status of treatments targeting FGFR, focusing on the trials that are evaluating the FGFR profile as inclusion criteria: Multi-Target, Pan-FGFR Inhibitors and anti-FGF (Fibroblast Growth Factor)/FGFR Monoclonal Antibodies. EXPERT OPINION: Most of the TKR share intracellular signaling pathways; therefore, cancer cells tend to overcome the inhibition of one tyrosine kinase receptor by activating another. The future of TKI (Tyrosine Kinase Inhibitor) therapy will potentially come from multi-targeted TKIs that target different TKR simultaneously. It is crucial to understand the interaction of the FGF-FGFR axis with other known driver TKRs. Based on this, it is possible to develop therapeutic strategies targeting multiple connected TKRs at once. One correct step in this direction is the reassessment of multi target inhibitors considering the FGFR status of the tumor. Another opportunity arises from assessing the use of FGFR TKI on patients harboring FGFR alterations.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC1cvps1ajsA%253D%253D&md5=d03086e258264a12c8c5cc7f2c5de8e3
15. 15
  Bunney, T. D.; Wan, S.; Thiyagarajan, N.; Sutto, L.; Williams, S. V.; Ashford, P.; Koss, H.; Knowles, M. a.; Gervasio, F. L.; Coveney, P. V.; Katan, M. The effect of mutations on drug sensitivity and kinase activity of fibroblast growth factor receptors: A combined experimental and theoretical study. EBioMedicine 2015, 2, 194– 204, DOI: 10.1016/j.ebiom.2015.02.009
  Google Scholar
  15
  The Effect of Mutations on Drug Sensitivity and Kinase Activity of Fibroblast Growth Factor Receptors: A Combined Experimental and Theoretical Study
  Bunney Tom D; Thiyagarajan Nethaji; Ashford Paul; Katan Matilda; Wan Shunzhou; Coveney Peter V; Sutto Ludovico; Gervasio Francesco L; Williams Sarah V; Knowles Margaret A; Koss Hans
  EBioMedicine (2015), 2 (3), 194-204 ISSN:.
  Fibroblast growth factor receptors (FGFRs) are recognized therapeutic targets in cancer. We here describe insights underpinning the impact of mutations on FGFR1 and FGFR3 kinase activity and drug efficacy, using a combination of computational calculations and experimental approaches including cellular studies, X-ray crystallography and biophysical and biochemical measurements. Our findings reveal that some of the tested compounds, in particular TKI258, could provide therapeutic opportunity not only for patients with primary alterations in FGFR but also for acquired resistance due to the gatekeeper mutation. The accuracy of the computational methodologies applied here shows a potential for their wider application in studies of drug binding and in assessments of functional and mechanistic impacts of mutations, thus assisting efforts in precision medicine.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC2srotlKmtA%253D%253D&md5=8414f3530d24bbba0ea8cc98cb784b96
16. 16
  Patani, H.; Bunney, T. D.; Thiyagarajan, N.; Norman, R. A.; Ogg, D.; Breed, J.; Ashford, P.; Potterton, A.; Edwards, M.; Katan, M. Landscape of activating cancer mutations in FGFR kinases and their differential responses to inhibitors in clinical use. Oncotarget 2016, 7, 24252– 24268, DOI: 10.18632/oncotarget.8132
  Google Scholar
  16
  Landscape of activating cancer mutations in FGFR kinases and their differential responses to inhibitors in clinical use
  Patani Harshnira; Bunney Tom D; Thiyagarajan Nethaji; Ashford Paul; Potterton Andrew; Edwards Mina; Pang Camilla S M; Orengo Christine; Katan Matilda; Norman Richard A; Ogg Derek; Breed Jason; Phillips Chris; Williams Sarah V; Knowles Margaret A; Thomson Gary S; Breeze Alexander L
  Oncotarget (2016), 7 (17), 24252-68 ISSN:.
  Frequent genetic alterations discovered in FGFRs and evidence implicating some as drivers in diverse tumors has been accompanied by rapid progress in targeting FGFRs for anticancer treatments. Wider assessment of the impact of genetic changes on the activation state and drug responses is needed to better link the genomic data and treatment options. We here apply a direct comparative and comprehensive analysis of FGFR3 kinase domain variants representing the diversity of point-mutations reported in this domain. We reinforce the importance of N540K and K650E and establish that not all highly activating mutations (for example R669G) occur at high-frequency and conversely, that some "hotspots" may not be linked to activation. Further structural characterization consolidates a mechanistic view of FGFR kinase activation and extends insights into drug binding. Importantly, using several inhibitors of particular clinical interest (AZD4547, BGJ-398, TKI258, JNJ42756493 and AP24534), we find that some activating mutations (including different replacements of the same residue) result in distinct changes in their efficacy. Considering that there is no approved inhibitor for anticancer treatments based on FGFR-targeting, this information will be immediately translatable to ongoing clinical trials.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A280%3ADC%252BC28fhsFShsw%253D%253D&md5=beb65006a52ef6062bab9c9d8086bf1d
17. 17
  Azam, M.; Seeliger, M. a.; Gray, N. S.; Kuriyan, J.; Daley, G. Q. Activation of tyrosine kinases by mutation of the gatekeeper threonine. Nat. Struct. Mol. Biol. 2008, 15, 1109– 1118, DOI: 10.1038/nsmb.1486
  Google Scholar
  17
  Activation of tyrosine kinases by mutation of the gatekeeper threonine
  Azam, Mohammad; Seeliger, Markus A.; Gray, Nathanael S.; Kuriyan, John; Daley, George Q.
  Nature Structural & Molecular Biology (2008), 15 (10), 1109-1118CODEN: NSMBCU; ISSN:1545-9993. (Nature Publishing Group)
  Protein kinases targeted by small-mol. inhibitors develop resistance through mutation of the 'gatekeeper' threonine residue of the active site. Here we show that the gatekeeper mutation in the cellular forms of c-ABL, c-SRC, platelet-derived growth factor receptor-α and -β, and epidermal growth factor receptor activates the kinase and promotes malignant transformation of BaF3 cells. Structural anal. reveals that a network of hydrophobic interactions - the hydrophobic spine - characteristic of the active kinase conformation is stabilized by the gatekeeper substitution. Substitution of glycine for the residues constituting the spine disrupts the hydrophobic connectivity and inactivates the kinase. Furthermore, a small-mol. inhibitor (compd. 14) that maximizes complementarity with the dismantled spine inhibits the gatekeeper mutation of BCR-ABL-T315I. These results demonstrate that mutation of the gatekeeper threonine is a common mechanism of activation for tyrosine kinases and provide structural insights to guide the development of next-generation inhibitors.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD1cXht1Sisb%252FM&md5=4afc4e5a0850a4510ced3fa5fbda97fa
18. 18
  van Gunsteren, W. F.; Daura, X.; Hansen, N.; Mark, A. E.; Oostenbrink, C.; Riniker, S.; Smith, L. J. Validation of Molecular Simulation: An Overview of Issues. Angew. Chem., Int. Ed. 2018, 57, 884– 902, DOI: 10.1002/anie.201702945
  Google Scholar
  18
  Validation of Molecular Simulation: An Overview of Issues
  van Gunsteren, Wilfred F.; Daura, Xavier; Hansen, Niels; Mark, Alan E.; Oostenbrink, Chris; Riniker, Sereina; Smith, Lorna J.
  Angewandte Chemie, International Edition (2018), 57 (4), 884-902CODEN: ACIEF5; ISSN:1433-7851. (Wiley-VCH Verlag GmbH & Co. KGaA)
  Computer simulation of mol. systems enables structure-energy-function relationships of mol. processes to be described at the sub-at., at., supra-at., or supra-mol. level. To interpret results of such simulations appropriately, the quality of the calcd. properties must be evaluated. This depends on the way the simulations are performed and on the way they are validated by comparison to values Qexp of exptl. observable quantities Q. One must consider (1) the accuracy of Qexp, (2) the accuracy of the function Q(rN) used to calc. a Q-value based on a mol. configuration rN of N particles, (3) the sensitivity of the function Q(rN) to the configuration rN, (4) the relative time scales of the simulation and expt., (5) the degree to which the calcd. and exptl. properties are equiv., and (6) the degree to which the system simulated matches the exptl. conditions. Exptl. data is limited in scope and generally corresponds to avs. over both time and space. A crit. anal. of the various factors influencing the apparent degree of (dis)agreement between simulations and expt. is presented and illustrated using examples from the literature. What can be done to enhance the validation of mol. simulation is also discussed.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXhs1ertQ%253D%253D&md5=d38b0f3fd0054e0fab46023a2c597442
19. 19
  Phillips, J. C.; Braun, R.; Wang, W.; Gumbart, J.; Tajkhorshid, E.; Villa, E.; Chipot, C.; Skeel, R. D.; Kal, L.; Schulten, K. Scalable molecular dynamics with NAMD. J. Comput. Chem. 2005, 26, 1781– 1802, DOI: 10.1002/jcc.20289
  Google Scholar
  19
  Scalable molecular dynamics with NAMD
  Phillips, James C.; Braun, Rosemary; Wang, Wei; Gumbart, James; Tajkhorshid, Emad; Villa, Elizabeth; Chipot, Christophe; Skeel, Robert D.; Kale, Laxmikant; Schulten, Klaus
  Journal of Computational Chemistry (2005), 26 (16), 1781-1802CODEN: JCCHDD; ISSN:0192-8651. (John Wiley & Sons, Inc.)
  NAMD is a parallel mol. dynamics code designed for high-performance simulation of large biomol. systems. NAMD scales to hundreds of processors on high-end parallel platforms, as well as tens of processors on low-cost commodity clusters, and also runs on individual desktop and laptop computers. NAMD works with AMBER and CHARMM potential functions, parameters, and file formats. This article, directed to novices as well as experts, first introduces concepts and methods used in the NAMD program, describing the classical mol. dynamics force field, equations of motion, and integration methods along with the efficient electrostatics evaluation algorithms employed and temp. and pressure controls used. Features for steering the simulation across barriers and for calcg. both alchem. and conformational free energy differences are presented. The motivations for and a roadmap to the internal design of NAMD, implemented in C++ and based on Charm++ parallel objects, are outlined. The factors affecting the serial and parallel performance of a simulation are discussed. Finally, typical NAMD use is illustrated with representative applications to a small, a medium, and a large biomol. system, highlighting particular features of NAMD, for example, the Tcl scripting language. The article also provides a list of the key features of NAMD and discusses the benefits of combining NAMD with the mol. graphics/sequence anal. software VMD and the grid computing/collab. software BioCoRE. NAMD is distributed free of charge with source code at www.ks.uiuc.edu.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2MXht1SlsbbJ&md5=189051128443b547f4300a1b8fb0e034
20. 20
  Coveney, P. V.; Wan, S. On the calculation of equilibrium thermodynamic properties from molecular dynamics. Phys. Chem. Chem. Phys. 2016, 18, 30236– 30240, DOI: 10.1039/C6CP02349E
  Google Scholar
  20
  On the calculation of equilibrium thermodynamic properties from molecular dynamics
  Coveney, Peter V.; Wan, Shunzhou
  Physical Chemistry Chemical Physics (2016), 18 (44), 30236-30240CODEN: PPCPFQ; ISSN:1463-9076. (Royal Society of Chemistry)
  The purpose of statistical mechanics is to provide a route to the calcn. of macroscopic properties of matter from their constituent microscopic components. It is well known that the macrostates emerge as ensemble avs. of microstates. However, this is more often stated than implemented in computer simulation studies. Here we consider foundational aspects of statistical mechanics which are overlooked in most textbooks and research articles that purport to compute macroscopic behavior from microscopic descriptions based on classical mechanics and show how due attention to these issues leads in directions which have not been widely appreciated in the field of mol. dynamics simulation.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XntVyisr8%253D&md5=f7950af8255cf494a117d46bc1b5191e
21. 21
  Lim, N. M.; Wang, L.; Abel, R.; Mobley, D. L. Sensitivity in binding free energies due to protein reorganization. J. Chem. Theory Comput. 2016, 12, 4620– 4631, DOI: 10.1021/acs.jctc.6b00532
  Google Scholar
  21
  Sensitivity in Binding Free Energies Due to Protein Reorganization
  Lim, Nathan M.; Wang, Lingle; Abel, Robert; Mobley, David L.
  Journal of Chemical Theory and Computation (2016), 12 (9), 4620-4631CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
  Tremendous recent improvements in computer hardware, coupled with advances in sampling techniques and force fields, are now allowing protein-ligand binding free energy calcns. to be routinely used to aid pharmaceutical drug discovery projects. However, despite these recent innovations, there are still needs for further improvement in sampling algorithms to more adequately sample protein motion relevant to protein-ligand binding. Here, we report our work identifying and studying such clear and remaining needs in the apolar cavity of T4 lysozyme L99A. In this study, we model recent exptl. results that show the progressive opening of the binding pocket in response to a series of homologous ligands. Even while using enhanced sampling techniques, we demonstrate that the predicted relative binding free energies (RBFE) are sensitive to the initial protein conformational state. Particularly, we highlight the importance of sufficient sampling of protein conformational changes and demonstrate how inclusion of three key protein residues in the "hot" region of the FEP/REST simulation improves the sampling and resolves this sensitivity, given enough simulation time.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28Xht1CltrbO&md5=5f4525fb7a55b6c9ad0510168e13f316
22. 22
  Clark, A. J.; Gindin, T.; Zhang, B.; Wang, L.; Abel, R.; Murret, C. S.; Xu, F.; Bao, A.; Lu, N. J.; Zhou, T.; Kwong, P. D.; Shapiro, L.; Honig, B.; Friesner, R. A. Free energy perturbation calculation of relative binding free energy between broadly neutralizing antibodies and the gp120 glycoprotein of HIV-1. J. Mol. Biol. 2017, 429, 930– 947, DOI: 10.1016/j.jmb.2016.11.021
  Google Scholar
  22
  Free energy perturbation calculation of relative binding free energy between broadly neutralizing antibodies and the gp120 glycoprotein of HIV-1
  Clark, Anthony J.; Gindin, Tatyana; Zhang, Baoshan; Wang, Lingle; Abel, Robert; Murret, Colleen S.; Xu, Fang; Bao, Amy; Lu, Nina J.; Zhou, Tongqing; Kwong, Peter D.; Shapiro, Lawrence; Honig, Barry; Friesner, Richard A.
  Journal of Molecular Biology (2017), 429 (7), 930-947CODEN: JMOBAK; ISSN:0022-2836. (Elsevier Ltd.)
  Direct calcn. of relative binding affinities between antibodies and antigens is a long-sought goal. However, despite substantial efforts, no generally applicable computational method has been described. Here, we describe a systematic free energy perturbation (FEP) protocol and calc. the binding affinities between the gp120 envelope glycoprotein of HIV-1 and three broadly neutralizing antibodies (bNAbs) of the VRC01 class. The protocol has been adapted from successful studies of small mols. to address the challenges assocd. with modeling protein-protein interactions. Specifically, we built homol. models of the three antibody-gp120 complexes, extended the sampling times for large bulky residues, incorporated the modeling of glycans on the surface of gp120, and utilized continuum solvent-based loop prediction protocols to improve sampling. We present three exptl. surface plasmon resonance data sets, in which antibody residues in the antibody/gp120 interface were systematically mutated to alanine. The RMS error in the large set (55 total cases) of FEP tests as compared to these expts., 0.68 kcal/mol, is near exptl. accuracy, and it compares favorably with the results obtained from a simpler, empirical methodol. The correlation coeff. for the combined data set including residues with glycan contacts, R2 = 0.49, should be sufficient to guide the choice of residues for antibody optimization projects, assuming that this level of accuracy can be realized in prospective prediction. More generally, these results are encouraging with regard to the possibility of using an FEP approach to calc. the magnitude of protein-protein binding affinities.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC28XitVSmsL3I&md5=ff264d642726916a45bae5b121bbdd06
23. 23
  Fratev, F.; Sirimulla, S. An improved free energy perturbation FEP+ sampling protocol for flexible ligand-binding domains. ChemRxiv 2018, DOI: 10.26434/chemrxiv.6204167.v1 .
  Google Scholar
  There is no corresponding record for this reference.
24. 24
  Fiser, A.; Sali, A. ModLoop: automated modeling of loops in protein structures. Bioinformatics 2003, 19, 2500– 2501, DOI: 10.1093/bioinformatics/btg362
  Google Scholar
  24
  ModLoop: automated modeling of loops in protein structures
  Fiser, Andras; Sali, Andrej
  Bioinformatics (2003), 19 (18), 2500-2501CODEN: BOINFP; ISSN:1367-4803. (Oxford University Press)
  Summary: ModLoop is a web server for automated modeling of loops in protein structures. The input is the at. coordinates of the protein structure in the Protein Data Bank format, and the specification of the starting and ending residues of one or more segments to be modeled, contg. no more than 20 residues in total. The output is the coordinates of the non-hydrogen atoms in the modeled segments. A user provides the input to the server via a simple web interface, and receives the output by e-mail. The server relies on the loop modeling routine in MODELLER that predicts the loop conformations by satisfaction of spatial restraints, without relying on a database of known protein structures. For a rapid response, ModLoop runs on a cluster of Linux PC computers.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD3sXpvVSisLk%253D&md5=4b0fdaca0a412715682496883ab2019c
25. 25
  Sohl, C. D.; Ryan, M. R.; Luo, B.; Frey, K. M.; Anderson, K. S. Illuminating the molecular mechanisms of tyrosine kinase inhibitor resistance for the FGFR1 gatekeeper mutation: the Achilles’ heel of targeted therapy. ACS Chem. Biol. 2015, 10, 1319– 1329, DOI: 10.1021/acschembio.5b00014
  Google Scholar
  25
  Illuminating the Molecular Mechanisms of Tyrosine Kinase Inhibitor Resistance for the FGFR1 Gatekeeper Mutation: The Achilles' Heel of Targeted Therapy
  Sohl, Christal D.; Ryan, Molly R.; Luo, BeiBei; Frey, Kathleen M.; Anderson, Karen S.
  ACS Chemical Biology (2015), 10 (5), 1319-1329CODEN: ACBCCT; ISSN:1554-8929. (American Chemical Society)
  Human fibroblast growth factor receptors (FGFRs) 1-4 are a family of receptor tyrosine kinases that can serve as drivers of tumorigenesis. In particular, FGFR1 gene amplification has been implicated in squamous cell lung and breast cancers. Tyrosine kinase inhibitors (TKIs) targeting FGFR1, including AZD4547 and E3810 (Lucitanib), are currently in early phase clin. trials. Unfortunately, drug resistance limits the long-term success of TKIs, with mutations at the "gatekeeper" residue leading to tumor progression. Here we show the first structural and kinetic characterization of the FGFR1 gatekeeper mutation, V561M FGFR1. The V561M mutation confers a 38-fold increase in autophosphorylation achieved at least in part by a network of interacting residues forming a hydrophobic spine to stabilize the active conformation. Moreover, kinetic assays established that the V561M mutation confers significant resistance to E3810, while retaining affinity for AZD4547. Structural analyses of these TKIs with wild type (WT) and gatekeeper mutant forms of FGFR1 offer clues to developing inhibitors that maintain potency against gatekeeper mutations. We show that AZD4547 affinity is preserved by V561M FGFR1 due to a flexible linker that allows multiple inhibitor binding modes. This is the first example of a TKI binding in distinct conformations to WT and gatekeeper mutant forms of FGFR, highlighting adaptable regions in both the inhibitor and binding pocket crucial for drug design. Exploiting inhibitor flexibility to overcome drug resistance has been a successful strategy for combating diseases such as AIDS and may be an important approach for designing inhibitors effective against kinase gatekeeper mutations.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXislGitLo%253D&md5=6c13f193e9e4c3f38b1e59f46fbc6342
26. 26
  Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Montgomery, J. A., Jr.; Vreven, T.; Kudin, K. N.; Burant, J. C.; Millam, J. M.; Iyengar, S. S.; Tomasi, J.; Barone, V.; Mennucci, B.; Cossi, M.; Scalmani, G.; Rega, N.; Petersson, G. A.; Nakatsuji, H.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Klene, M.; Li, X.; Knox, J. E.; Hratchian, H. P.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Ayala, P. Y.; Morokuma, K.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Zakrzewski, V. G.; Dapprich, S.; Daniels, A. D.; Strain, M. C.; Farkas, O.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Ortiz, J. V.; Cui, Q.; Baboul, A. G.; Clifford, S.; Cioslowski, J.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Gonzalez, C.; Pople, J. A. Gaussian 03; Gaussian, Inc.: Wallingford, CT, 2004.
  Google Scholar
  There is no corresponding record for this reference.
27. 27
  Case, D. A.; Cheatham, T. E.; Darden, T.; Gohlke, H.; Luo, R.; Merz, K. M.; Onufriev, A.; Simmerling, C.; Wang, B.; Woods, R. J. The Amber biomolecular simulation programs. J. Comput. Chem. 2005, 26, 1668– 1688, DOI: 10.1002/jcc.20290
  Google Scholar
  27
  The amber biomolecular simulation programs
  Case, David A.; Cheatham, Thomas E., III; Darden, Tom; Gohlke, Holger; Luo, Ray; Merz, Kenneth M., Jr.; Onufriev, Alexey; Simmerling, Carlos; Wang, Bing; Woods, Robert J.
  Journal of Computational Chemistry (2005), 26 (16), 1668-1688CODEN: JCCHDD; ISSN:0192-8651. (John Wiley & Sons, Inc.)
  The authors describe the development, current features, and some directions for future development of the Amber package of computer programs. This package evolved from a program that was constructed in the late 1970s to do Assisted Model Building with Energy Refinement, and now contains a group of programs embodying a no. of powerful tools of modern computational chem., focused on mol. dynamics and free energy calcns. of proteins, nucleic acids, and carbohydrates.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2MXht1SlsbbM&md5=93be29ff894bab96c783d24e9886c7d0
28. 28
  Wang, J.; Wolf, R. M.; Caldwell, J. W.; Kollman, P. A.; Case, D. A. Development and testing of a general amber force field. J. Comput. Chem. 2004, 25, 1157– 1174, DOI: 10.1002/jcc.20035
  Google Scholar
  28
  Development and testing of a general Amber force field
  Wang, Junmei; Wolf, Romain M.; Caldwell, James W.; Kollman, Peter A.; Case, David A.
  Journal of Computational Chemistry (2004), 25 (9), 1157-1174CODEN: JCCHDD; ISSN:0192-8651. (John Wiley & Sons, Inc.)
  We describe here a general Amber force field (GAFF) for org. mols. GAFF is designed to be compatible with existing Amber force fields for proteins and nucleic acids, and has parameters for most org. and pharmaceutical mols. that are composed of H, C, N, O, S, P, and halogens. It uses a simple functional form and a limited no. of atom types, but incorporates both empirical and heuristic models to est. force consts. and partial at. charges. The performance of GAFF in test cases is encouraging. In test I, 74 crystallog. structures were compared to GAFF minimized structures, with a root-mean-square displacement of 0.26 Å, which is comparable to that of the Tripos 5.2 force field (0.25 Å) and better than those of MMFF 94 and CHARMm (0.47 and 0.44 Å, resp.). In test II, gas phase minimizations were performed on 22 nucleic acid base pairs, and the minimized structures and intermol. energies were compared to MP2/6-31G* results. The RMS of displacements and relative energies were 0.25 Å and 1.2 kcal/mol, resp. These data are comparable to results from Parm99/RESP (0.16 Å and 1.18 kcal/mol, resp.), which were parameterized to these base pairs. Test III looked at the relative energies of 71 conformational pairs that were used in development of the Parm99 force field. The RMS error in relative energies (compared to expt.) is about 0.5 kcal/mol. GAFF can be applied to wide range of mols. in an automatic fashion, making it suitable for rational drug design and database searching.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BD2cXksFakurc%253D&md5=2992017a8cf51f89290ae2562403b115
29. 29
  Maier, J. A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.; Hauser, K. E.; Simmerling, C. ff14SB: Improving the accuracy of protein side chain and backbone parameters from ff99SB. J. Chem. Theory Comput. 2015, 11, 3696– 3713, DOI: 10.1021/acs.jctc.5b00255
  Google Scholar
  29
  ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB
  Maier, James A.; Martinez, Carmenza; Kasavajhala, Koushik; Wickstrom, Lauren; Hauser, Kevin E.; Simmerling, Carlos
  Journal of Chemical Theory and Computation (2015), 11 (8), 3696-3713CODEN: JCTCCE; ISSN:1549-9618. (American Chemical Society)
  Mol. mechanics is powerful for its speed in atomistic simulations, but an accurate force field is required. The Amber ff99SB force field improved protein secondary structure balance and dynamics from earlier force fields like ff99, but weaknesses in side chain rotamer and backbone secondary structure preferences have been identified. Here, we performed a complete refit of all amino acid side chain dihedral parameters, which had been carried over from ff94. The training set of conformations included multidimensional dihedral scans designed to improve transferability of the parameters. Improvement in all amino acids was obtained as compared to ff99SB. Parameters were also generated for alternate protonation states of ionizable side chains. Av. errors in relative energies of pairs of conformations were under 1.0 kcal/mol as compared to QM, reduced 35% from ff99SB. We also took the opportunity to make empirical adjustments to the protein backbone dihedral parameters as compared to ff99SB. Multiple small adjustments of φ and ψ parameters were tested against NMR scalar coupling data and secondary structure content for short peptides. The best results were obtained from a phys. motivated adjustment to the φ rotational profile that compensates for lack of ff99SB QM training data in the β-ppII transition region. Together, these backbone and side chain modifications (hereafter called ff14SB) not only better reproduced their benchmarks, but also improved secondary structure content in small peptides and reprodn. of NMR χ1 scalar coupling measurements for proteins in soln. We also discuss the Amber ff12SB parameter set, a preliminary version of ff14SB that includes most of its improvements.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhtFequ7rN&md5=7b803577b3b6912cc6750cfbd356596e
30. 30
  Jo, S.; Jiang, W. A generic implementation of replica exchange with solute tempering (REST2) algorithm in NAMD for complex biophysical simulations. Comput. Phys. Commun. 2015, 197, 304– 311, DOI: 10.1016/j.cpc.2015.08.030
  Google Scholar
  30
  A generic implementation of replica exchange with solute tempering (REST2) algorithm in NAMD for complex biophysical simulations
  Jo, Sunhwan; Jiang, Wei
  Computer Physics Communications (2015), 197 (), 304-311CODEN: CPHCBZ; ISSN:0010-4655. (Elsevier B.V.)
  Replica Exchange with Solute Tempering (REST2) is a powerful sampling enhancement algorithm of mol. dynamics (MD) in that it needs significantly smaller no. of replicas but achieves higher sampling efficiency relative to std. temp. exchange algorithm. In this paper, we extend the applicability of REST2 for quant. biophys. simulations through a robust and generic implementation in greatly scalable MD software NAMD. The rescaling procedure of force field parameters controlling REST2 "hot region" is implemented into NAMD at the source code level. A user can conveniently select hot region through VMD and write the selection information into a PDB file. The rescaling keyword/parameter is written in NAMD Tcl script interface that enables an on-the-fly simulation parameter change. Our implementation of REST2 is within communication-enabled Tcl script built on top of Charm++, thus communication overhead of an exchange attempt is vanishingly small. Such a generic implementation facilitates seamless cooperation between REST2 and other modules of NAMD to provide enhanced sampling for complex biomol. simulations. Three challenging applications including native REST2 simulation for peptide folding-unfolding transition, free energy perturbation/REST2 for abs. binding affinity of protein-ligand complex and umbrella sampling/REST2 Hamiltonian exchange for free energy landscape calcn. were carried out on IBM Blue Gene/Q supercomputer to demonstrate efficacy of REST2 based on the present implementation.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC2MXhsFSlu73I&md5=776ea438c536bf84540b1443494e647c
31. 31
  Ballester, P. J.; Mitchell, J. B. O. A machine learning approach to predicting protein–ligand binding affinity with applications to molecular docking. Bioinformatics 2010, 26, 1169– 1175, DOI: 10.1093/bioinformatics/btq112
  Google Scholar
  31
  A machine learning approach to predicting protein-ligand binding affinity with applications to molecular docking
  Ballester, Pedro J.; Mitchell, John B. O.
  Bioinformatics (2010), 26 (9), 1169-1175CODEN: BOINFP; ISSN:1367-4803. (Oxford University Press)
  Accurately predicting the binding affinities of large sets of diverse protein-ligand complexes is an extremely challenging task. The scoring functions that attempt such computational prediction are essential for analyzing the outputs of mol. docking, which in turn is an important technique for drug discovery, chem. biol. and structural biol. Each scoring function assumes a predetd. theory-inspired functional form for the relationship between the variables that characterize the complex, which also include parameters fitted to exptl. or simulation data and its predicted binding affinity. The inherent problem of this rigid approach is that it leads to poor predictivity for those complexes that do not conform to the modeling assumptions. Moreover, resampling strategies, such as cross-validation or bootstrapping, are still not systematically used to guard against the overfitting of calibration data in parameter estn. for scoring functions. We propose a novel scoring function (RF-Score) that circumvents the need for problematic modeling assumptions via non-parametric machine learning. In particular, Random Forest was used to implicitly capture binding effects that are hard to model explicitly. RF-Score is compared with the state of the art on the demanding PDBbind benchmark. Results show that RF-Score is a very competitive scoring function. Importantly, RF-Score's performance was shown to improve dramatically with training set size and hence the future availability of more high-quality structural and interaction data is expected to lead to improved versions of RF-Score.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC3cXlt1Cjs78%253D&md5=6d2d48344dc5ba70f0ff85f432d0ec63
32. 32
  Jimenez, J.; Skalic, M.; Martinez-Rosell, G.; De Fabritiis, G. KDEEP: Protein–Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks. J. Chem. Inf. Model. 2018, 58, 287– 296, DOI: 10.1021/acs.jcim.7b00650
  Google Scholar
  32
  KDEEP: Protein-Ligand Absolute Binding Affinity Prediction via 3D-Convolutional Neural Networks
  Jimenez, Jose; Skalic, Miha; Martinez-Rosell, Gerard; De Fabritiis, Gianni
  Journal of Chemical Information and Modeling (2018), 58 (2), 287-296CODEN: JCISD8; ISSN:1549-9596. (American Chemical Society)
  Accurately predicting protein-ligand binding affinities is an important problem in computational chem. since it can substantially accelerate drug discovery for virtual screening and lead optimization. We propose here a fast machine-learning approach for predicting binding affinities using state-of-the-art 3D-convolutional neural networks and compare this approach to other machine-learning and scoring methods using several diverse data sets. The results for the std. PDBbind (v.2016) core test-set are state-of-the-art with a Pearson's correlation coeff. of 0.82 and a RMSE of 1.27 in pK units between exptl. and predicted affinity, but accuracy is still very sensitive to the specific protein used. KDEEP is made available via PlayMol.org for users to test easily their own protein-ligand complexes, with each prediction taking a fraction of a second. We believe that the speed, performance, and ease of use of KDEEP makes it already an attractive scoring function for modern computational chem. pipelines.
  >> More from SciFinder ^®
  https://chemport.cas.org/services/resolver?origin=ACS&resolution=options&coi=1%3ACAS%3A528%3ADC%252BC1cXkslSitQ%253D%253D&md5=81943a6732be99e5439e1e3a25fa4414
33. 33
  Coveney, P. V.; Dougherty, E. R.; Highfield, R. R. Big data need big theory too. Philos. Trans. R. Soc., A 2016, 374, 20160153, DOI: 10.1098/rsta.2016.0153
  Google Scholar
  There is no corresponding record for this reference.
Supporting Information
Supporting Information
ARTICLE SECTIONS
Jump To
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jctc.8b01118.
- Tables containing individual ΔG_alch values for all the relative free energy calculations, and a table for the comparison of occupancies of hydrogen bonds between the inhibitors and the protein (PDF)
- Topology and coordinate files for all ligand–protein complexes (ZIP)
- ct8b01118_si_001.pdf (193.8 kb)
- ct8b01118_si_002.zip (40.33 MB)
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.

PDF [ 4MB]

All Types

Ensemble-Based Replica Exchange Alchemical Free Energy Methods: The Effect of Protein Mutations on Inhibitor Binding

Article Views

Altmetric

Citations

Abstract

1. Introduction

Figure 1

2. Methods

2.1. Hybrid Topology

Figure 2

2.2. Free Energy Schemes

2.3. REST2 Region

2.4. Simulation Setup

2.5. Simulations

2.6. Computational Resources

3. Results

3.1. Local Mutation

Figure 3

3.2. Remote Mutations

Figure 4

3.3. Effect of Extending Simulation Time

Figure 5

Figure 6

4. Discussion

4.1. Improved Sampling of AZD4547 on “Heating”

Figure 7

Figure 8

Figure 9

4.2. The Exceptional Case of TKI258: Limitations of λ-REST2

Figure 10

Figure 11

Figure 12

Figure 13

5. Conclusions

Supporting Information

Terms & Conditions

Author Information

Acknowledgments

References

Cited By

Abstract

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

References

Supporting Information

Supporting Information

Terms & Conditions

STEP 1:

STEP 2: