ACS Editors’ Choice

Transthyretin (TTR) plays a vital role in thyroid hormone transport and homeostasis in both the blood and target tissues. Interactions between exogenous compounds and TTR can disrupt the function of the endocrine system, potentially causing toxicity. In the Tox24 challenge, we leveraged the data set provided by the organizers to develop a deep learning-based consensus model, integrating sPhysNet, KANO, and GGAP-CPI for predicting TTR binding affinity. Each model utilized distinct levels of molecular information, including 2D topology, 3D geometry, and protein–ligand interactions. Our consensus model achieved favorable performance on the blind test set, yielding an RMSE of 20.8 and ranking fifth among all submissions. Following the release of the blind test set, we incorporated the leaderboard test set into our training data, further reducing the RMSE to 20.6 in an offlineretrospective study. These results demonstrate that combining three regression models across different modalities significantly enhances the predictive accuracy. Furthermore, we employ the standard deviation of the consensus model’s ensemble outputs as an uncertainty estimate. Our analysis reveals that both the RMSE and interval error of predictions increase with rising uncertainty, indicating that the uncertainty can serve as a useful measure of prediction confidence. We believe that this consensus model can be a valuable resource for identifying potential TTR binders and predicting their binding affinity in silico. The source code for data preparation, model training, and prediction can be accessed at https://github.com/xiaolinpan/tox24_challenge_submission_yingkai_lab.

Aldo-keto reductase family 1 member C3 (AKR1C3) is a member of the AKR superfamily of enzymes that metabolize androgen, estrogen, and prostaglandin substrates that drive proliferation in hormone-dependent cancers. Interest in developing selective inhibitors has produced tool compounds for the inactivation or degradation of AKR1C3 with varying degrees of selectivity among the 14 known AKR proteins. Selectivity of AKR1C3 inhibitors across the AKR family is critical since a clinical candidate failed due to hepatotoxicity from off-target inhibition of AKR1D1. Here, we report development of a sulfonyl-triazole (SuTEx) covalent AKR1C3 inhibitor (RJG-2051) that selectively engages a noncatalytic tyrosine residue (Y24) on AKR1C3. Importantly, RJG-2051 exhibited negligible cross-reactivity with AKRs or other proteins across 1800+ tyrosine and lysine sites quantified by chemical proteomics. Our disclosure of a covalent inhibitor for potent AKR1C3 inactivation with proteome-wide selectivity in cells will expedite cell biological studies for testing the therapeutic potential of this metabolic target.

Hydrogel-based drug delivery systems hold significant clinical potential by enabling precise spatial and temporal control over therapeutic release, ranging from metabolites, macromolecules to other cellular and subcellular constructs. However, achieving programmable release of payloads with diverse molecular weights at distinct rates typically requires complex polymer designs that can compromise the accessibility and biocompatibility of the delivery system. We present a scalable method for producing injectable, micrometer-scale alginate hydrogel particles (microgels) with precisely tuned microstructures for multiplexed, programmable cargo release. Our approach integrates an established jetting technique with a simple postsynthesis ion-exchange process to fine-tune the cross-linked microstructure of alginate microgels. By varying cation type (Ca²⁺, Mg²⁺, Na⁺) and concentration, we systematically modulate the microgels’ chemical and physical properties to control release rates of model compounds, including rhodamine B, methylene blue, and dextrans of various molecular weights. Additionally, a PEG-alginate composite microgel system is used to demonstrate the pre-programmed stepwise release of rhodamine B. These findings offer a straightforward strategy for postsynthetic manipulation of ionic microgels with controllable release performances, paving the way for advanced biomedical applications.

Schrock-type alkylidene complexes, especially imido ligand-supported metal alkylidene complexes, have been used as catalysts and reagents for C–C multiple bond-forming reactions involving unsaturated organic compounds. Herein, we report that an imidotungsten(IV) complex, WCl₂(=NDipp)(P(OMe)₃)₃, catalyzes a [2+2+1]-cycloaddition reaction of 3,3-disubstituted cyclopropenes and 2 equiv of alkynes, giving multisubstituted cyclopentadienes. The initial step of this catalytic reaction is ring-opening of a cyclopropene by the (imido)tungsten(IV) complex to form an (imido)tungsten vinylalkylidene species. Stoichiometric reactions of the (imido)tungsten vinylalkylidene complex and an η²-alkyne tungsten complex as well as density functional theory (DFT) calculations revealed the reaction mechanism, in which two consecutive metatheses of the (imido)tungsten vinylalkylidene species and two alkynes followed by reductive elimination afford the multisubstituted cyclopentadiene along with a regeneration of an (imido)tungsten(IV) species.

Compared to Danishefsky’s diene, Rawal’s diene suffers from low commercial availability, limited scalability, and obscure stability issues, which strictly limits its usability. Herein, we present an optimized, scalable synthetic protocol that achieves yields suitable for the diene’s semi-industrial production, with adjustments to reagent concentrations, reaction conditions, and isolation procedures to enhance efficiency. Complementing synthetic advancements, this work explores the diene’s physicochemical stability under diverse storage conditions. Rigorous quality control methodologies exploiting nuclear magnetic resonance (NMR) and infrared (IR) spectroscopy facilitate precise monitoring of purity and degradation pathways, establishing robust analytical standards. Additionally, the work demonstrates the utility and advantages of Rawal’s diene in multigram syntheses of 2-alkyl-2,3-dihydro-4H-pyran-4-ones, showcasing its applicability for medicinal chemistry purposes. The findings disclosed in the paper establish a foundation for the broader adoption and commercialization of Rawal’s diene, enabling its integration into academic and industrial workflows.

The possibility of using light to image and manipulate neuronal activity, at the heart of Neurophotonics, has provided new irreplaceable tools to study brain function. In particular, the combination of multiphoton microscopy and optogenetics allows researchers to interact with neuronal circuits with single-cell resolution in living brain tissues. However, significant optical challenges remain to empower new discoveries in Neuroscience. This Review focuses on three critical areas for future development: (1) expanding imaging and optogenetic stimulation to larger fields of view and faster acquisition speeds, while maintaining single-cell resolution and minimizing photodamage; (2) enabling access to deeper brain regions to study currently inaccessible neuronal circuits; and (3) developing optical techniques for studying natural behaviors in freely moving animals. For each of these challenges, we review the current state-of-the-art and suggest future directions with the potential to transform the field.

Predicting the water phase diagram is a powerful way to evaluate water models through molecular simulations. Equilibrium points are usually obtained through free energy calculations or direct coexistence simulations in the NPT ensemble. The former can be complex, especially for ice with partial proton order, while the latter can require multiple long and computationally costly simulations. In this work, we report the melting points of ice Ih, III, V, and VI between 0.1 and 1190 MPa through molecular dynamics direct coexistence simulations in the NPH ensemble. Our results are consistent with the original TIP4P/Ice work coexistence lines, except for ice III, for which we report a much larger stability region. Our data agree with recent works, validating this methodology as an alternative to free energy calculations and NPT direct coexistence simulations for high-pressure phases of ice.

With the emergence of generative artificial intelligence (GenAI) and ChatGPT, individualization of learning pathways was discussed in the educational context. One way to achieve customization is the use of specialized versions of ChatGPT–so-called GPTs. Those GPTs are publicly available and used daily by thousands of people worldwide. In this work, we analyze the current market of chemistry-specialized GPTs in the OpenAI GPT ‘store’. We identify common shortcomings, such as unnecessary activation of features like DALL-E, overly positive use of terms like ‘expert’ or ‘teacher’ and the inclusion of copyrighted material. To highlight the urgency of well-designed GPTs in chemistry education, we analyze the most used chemistry-specific GPT in more detail to discuss further problems with GPTs. In summary, we advocate a collaborative effort within the chemistry education community to develop robust and publicly available GPTs. Future research might benefit from a focus on creating and refining these GPTs and exploring their impact on learning in both secondary and tertiary educational settings.

In this work, we analyze the current state of mechanistic understanding in CO₂ electrolysis and give our best analysis of what we believe is the most dominant mechanism for the predominant products in CO₂ electrolysis. We draw on both computational and experimental literature to develop conclusions for C1 and C2 products. From this, we develop a set of self-consistent mechanistic rules. As the volume of literature on the mechanism toward C3 products is substantially smaller than on C1 and C2 products, these rules help us in evaluating mechanistic pathways toward C3 products. While these mechanistic pathways are speculative, it does give us a point of reference that can be modified in the future based on further developments in the field.

Since the seminal works on thermoreversible covalent networks followed by the discovery of vitrimers by L. Leibler and co-workers in 2011, numerous chemistries and strategies have flourished to design covalent adaptable networks (CANs) thus opening a novel research field. Using reversible covalent bonds that have been known for decades in molecular chemistry, CANs combine both the rheological characteristics of thermosets (chemically cross-linked networks, insoluble and infusible) and those of thermoplastics (entangled polymer chains able to be dissolved and to flow above their glass transition temperature). The aim of this tutorial review is to provide polymer chemists with guidelines to precisely and properly characterize CANs. Depending on the nature of the exchange mechanism (dissociative, associative, or a combination of both), on the kinetics of exchange, and on the cross-link density, characteristic relaxation times can vary from less than a second to a few hours. The time scale and distribution of relaxation times influence the rheological experiments and models that should be used. The present didactic review provides, from the rich recent literature, a guideline for adequate material characterizations and rheological measurements (and theoretical models applicable) that have been used to study CAN viscoelastic and thermomechanical properties.

Ebola virus (EBOV) causes deadly Ebola virus disease (EVD), and EVD survivors are at high risk of developing blinding ocular complications, which associate with the breakdown of human retinal pigment epithelial (RPE) barrier. Here, we demonstrated EBOV glycoprotein (GP) could directly impair RPE barrier function. EBOV GP significantly decreased expression of tight junction (TJ) proteins in RPE monolayers, resulting in an increase of monolayer permeability. EBOV GP activated PI3K/Akt pathway and induced oxidative stress in RPE cells as evidenced by an increase in the production of reactive oxygen species (ROS), decreased expression of an antioxidant factor, nuclear factor erythroid 2-related factor 2 (Nrf2), and its downstream proteins heme oxygenase 1 (HO-1) and NAD(P)H dehydrogenase (quinone) 1 (NQO1). We found activating Nrf2 could counteract EBOV GP-induced RPE barrier injury. Furthermore, GP2 subunit is the key region in the GP that impairs RPE barrier function. Destruction of RPE barrier function by EBOV GP leads to translocation of bacteria and HIV-1. We confirmed EBOV GP-mediated impairment of RPE barrier function in mice. As an Nrf2 activator, resveratrol displays protective effects on the RPE barrier function. Collectively, our study demonstrates EBOV GP impairs the RPE barrier function through PI3K/Akt-Nrf2 pathway and resveratrol is a promising therapeutic agent for EVD-associated retinal complications.

Sleeping microenvironments (SMEs) can expose young children to chemicals of concern. Using passive samplers, we measured the concentrations of ortho-phthalates (PAEs), organophosphate esters (OPEs), and UV-filters (benzophenones, salicylates, and phenolic benzotriazoles) in the bedroom air, SME, and released from mattresses in 25 bedrooms of children aged 6 months to 4 years in Toronto and Ottawa, Canada. We detected 28, 31, and 30 compounds in bedroom air, SME air, and mattresses, respectively. SME exceeded bedroom air concentrations, indicating elevated exposure while sleeping and sources from SME contents, with two exceptions. Higher concentrations of two PAEs and five OPEs (including isomers) in mattress versus SME samplers indicated that mattresses were a source. Bedding items were likely sources of tris(2-butoxyethyl) phosphate (TBOEP) where SME concentrations were significantly higher than those in mattress samplers. Older mattresses had higher concentrations of di-2-ethylhexyl phthalate (DEHP) and benzyl butyl phthalate (BzBP). These results indicate children’s exposure to a range of chemicals of concern while sleeping, at higher concentrations than in their bedrooms. Practical steps to reduce exposure include limiting items in SMEs such as toys and frequently washing bedding. Also, these results should prompt stricter regulations and greater producer responsibility regarding harmful chemicals used in mattresses and SME articles.

Understanding the principles that govern actinide–ligand (An–L) bonding is essential for advancing practical applications in nuclear industry and environmental protection, as well as for deepening our fundamental knowledge of actinide chemistry. Modifying the symmetry or softness of the coordinating ligand, or altering the metal center, are common strategies to modulate the energy and orbital overlap in An–L interactions, driving both experimental and computational research efforts toward greater control over covalency. On the metal side, reducing the oxidation state causes the f- and d-orbitals to become more diffuse and destabilized. This not only enhances covalency when coordinated to suitable ligands but also opens the door to novel bonding modes via metal-to-ligand back-donation which despite their potential for advancing separation chemistry, remain largely underexplored. On the ligand side, symmetry plays a critical role in controlling the types of bonding modes. In this work, we demonstrate that variations in actinide oxidation state across the early actinide series can be used as a lever to selectively activate or suppress back-bonds. By selecting three ligands─allyl, cyclocumulene, and cyclopropene─each possessing symmetries conducive to δ and φ back-bond formation, we identified a previously elusive φ “head-to-head” back-bond. This interaction emerged as the strongest in uranium and protactinium diallyl complexes, surpassing the φ back-bonds observed in cyclooctatetraene (COT) systems. Additionally, an extension of the Dewar-Chatt-Duncanson model to f-elements is proposed. These findings not only advance our fundamental understanding of actinide bonding but also open new pathways for 5f-electrons-driven chemistries.

The investigation of low ceiling temperature (T_c) monomers is an active area of research in the field of polymer science to address modern challenges in waste and recycling. Many commendable contributions have been made to this field; however, a thorough survey of the literature has revealed common oversights in the calculations of the changes in enthalpy (ΔH_p) and entropy (ΔS_p) upon polymerization as well as T_c. This Perspective aims to clarify how to avoid these pitfalls, accurately calculate these values, and outline best practices for experimentally measuring these key thermodynamic parameters. In an era where researchers are increasingly reliant on the quality of data, especially for endeavors in machine learning and artificial intelligence, it is important to establish a unified approach for making these calculations to optimize precision and accuracy.

Electrochemical ironmaking can provide an energy efficient, zero-emissions alternative to traditional methods of ironmaking, but the scalability of low-temperature electrochemical cells may be constrained by reactor throughput and the availability of acceptable feedstocks. Electrodes directly converting solid iron-oxide particles to metal circumvent traditional mass-transport limitations but are sensitive to both the particle size and nanoscale morphology of reactants. The effect of these properties on reactor throughput has not been systematically studied at model electrowinning surfaces. Here, we have used size-controlled, homologous α-Fe₂O₃ particles to study how the nanoscale morphology of oxides influences the obtainable current density toward Fe metal and integrated these results in a technoeconomic model for alkaline iron electrowinning systems. Micron-scale α-Fe₂O₃ with nanoscale porosity can be used to form Fe at current densities commensurate with industrial water electrolysis (>0.6 A cm^–2) in the absence of external convection, providing a path to cost-competitive and scalable ironmaking using electrochemistry.

Understanding protein–drug complex structures is crucial for elucidating therapeutic mechanisms and side effects. Blind docking facilitates site identification but is hindered by computational complexity and imprecise scoring, causing ambiguity. Dipolar electron paramagnetic resonance (EPR) provides spin–spin distances but struggles to determine relative positions within complexes. We present a novel approach combining GPU-accelerated blind docking with EPR distance constraints to enhance binding site detection. Our algorithm uses a single EPR distance distribution to filter and validate docking results. Ligand poses from blind docking are clustered, filtered by expected distances, and refined through focused docking. To illustrate our approach, we investigated human serum albumin binding with porphyrin-based photosensitizers used in photodynamic therapy. Combining docking and EPR, we identified possible binding sites, demonstrating that EPR data significantly reduce possible configurations and provide experimentally validated information. This strategy produces a detailed map of photoligand binding sites, revealing that binding may occur away from standard albumin sites and often involves multiple locations. Furthermore, it overcomes key limitations of fluorescence-based methods, which are prone to misinterpretation in albumin studies due to non one-to-one donor–acceptor relationships. By resolving ambiguities in both blind docking and EPR, our framework provides a versatile platform for investigating EPR-active ligands.

Postoperative abdominal adhesion is a prevalent issue with high incidence rates, often resulting in complications such as bowel obstruction and infertility. Currently, poly(lactic acid) (PLA)-based anti-adhesion membranes are extensively used for the prevention of abdominal adhesions. However, these membranes necessitate suturing, which increases the risk of secondary injury. In this study, we present a Janus patch with asymmetric adhesion properties designed to prevent postoperative abdominal adhesions. The patch consists of two functional layers: an adhesive layer made of a poly(lactic acid-co-ethylethylene phosphate) copolymer, which achieves tissue adhesion via hydrophilicity, hydrogen bonding, and electrostatic interactions, and a non-adhesive layer composed of electrospun PLA membrane. We characterized various properties of the Janus patch, including its morphology, adhesive properties, and biocompatibility. Adhesive properties tests revealed that the adhesive layer of the Janus patch demonstrated superior adhesive capabilities on various tissues compared to the non-adhesive PLA layer. In vivo experiments indicated that the asymmetric adhesive properties of the Janus patch effectively prevent postoperative abdominal adhesions. This work highlights a promising approach for addressing the challenges associated with adhesion prevention and secondary injuries, paving the way for safer and more effective postoperative care.

Dimensional analysis is an algorithm currently in use in almost every chemistry classroom in the United States. Chemistry educators use this procedural tool in the classroom with the intention of providing students with a reliable method to solve many of the relatively simple math problems they encounter. The unintended consequence of using this algorithm is that dimensional analysis divorces conceptual from procedural knowledge and effectively produces students who are answer-getters instead of critical thinkers. Instruction that integrates both conceptual and procedural knowledge into curriculum improves students’ conceptual understanding of a topic while still equipping them with necessary skills to solve rote, mathematical problems. This study assesses the impacts of replacing instruction in dimensional analysis with instruction in proportional reasoning on students’ conceptual chemistry understanding. Results indicate that students who received instruction in proportional reasoning instead of dimensional analysis experienced significantly greater growth in the area of chemistry conceptual understanding when compared to their peers who received instruction in dimensional analysis.

Cavity-enhanced diamond color center qubits can be initialized, manipulated, entangled, and read individually with high fidelity, which makes them ideal for large-scale, modular quantum computers, quantum networks, and distributed quantum sensing systems. However, diamond’s unique material properties pose significant challenges in manufacturing nanophotonic devices, leading to fabrication-induced structural imperfections and inaccuracies in defect implantation, which hinder reproducibility, degrade optical properties and compromise the spatial coupling of color centers to small mode-volume cavities. A cavity design tolerant to fabrication imperfections─such as surface roughness, sidewall slant, and nonoptimal emitter positioning─can improve coupling efficiency while simplifying fabrication. To address this challenge, a deep learning-based optimization methodology is developed to enhance the fabrication error tolerance of nanophotonic devices. Convolutional neural networks (CNNs) are applied to promising designs, such as L2 and fishbone nanobeam cavities, predicting Q-factors at least one-million times faster than traditional finite-difference time-domain (FDTD) simulations, enabling efficient optimization of complex, high-dimensional parameter spaces. The CNNs achieve prediction errors below 3.99% and correlation coefficients up to 0.988. Optimized structures demonstrate a 52% reduction in Q-factor degradation, achieving quality factors of 5 × 10⁴ under real-world conditions and a 2-fold expansion in field distribution, enabling efficient coupling of nonoptimally positioned emitters. Compared to previous deep-learning optimization methods, this approach achieves twice the Q-factor performance in the presence of fabrication errors, significantly enhancing device robustness. Hence, this methodology enables scalable, high-yield manufacturing of robust nanophotonic devices, including the cavity-enhanced diamond quantum systems developed in this study.

The ingenious exploitation of nanosized spatial confinement for the creation of distinct and independently functional sites has spearheaded an alluring avenue for the meticulous manipulation of activity and selectivity within heterogeneous photocatalysis. Herein, we report the controllable design of a dual-defect hollow hierarchical ZnIn₂S_x@CeO_x catalyst with spatially decoupled redox active sites, enabling the superior selective ethanol oxidation coupling to 2,3-butanediol (the highest generation rate of 3.51 mmol·g^–1·h^–1 so far) and CO₂ reduction to syngas within one photoredox cycle. Mechanistic investigations reveal that the fast charge transfer channels, stemming from the defect-induced interfacial Ce–S bonds, and the internal electric field of the heterostructure, facilitate the separation of photogenerated charge carrier along the Z-scheme transfer path. Simultaneously, the hollow heterostructures partition the exterior space from the interior cavity, enabling spatial separation of photogenerated electrons and holes to drive their respective half-reactions within stable and isolated microenvironments. Consequently, such a strategy allows ethanol and CO₂ to be individually activated in separated spaces without interference, resulting in a robust, highly active and selective ethanol coupling product, along with tunable syngas formation. This work highlights the significant potential of leveraging a spatially confined nanostructure in steering selective artificial photoredox coupling synthesis.