Journal of Chemical Information and Modeling 最新文献

{"title":"PSCG-Net: A Multiscale Crystal Graph Neural Network for Accelerated Materials Discovery.","authors":"Guangyao Chen,Zhilong Wang,Fengqi You","doi":"10.1021/acs.jcim.5c01460","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01460","url":null,"abstract":"The discovery of new materials is crucial for progress in energy, electronics, and sustainable technology. Traditional machine learning approaches, including graph neural networks (GNNs), often fall short because they cannot capture long-range interactions in crystalline materials due to fixed cutoff radii. To overcome this limitation, a pair-scalable crystal graph neural network (PSCG-Net) is proposed. This framework incorporates multiscale structural representations inspired by the pair distribution function and uses graphs with various cutoff distances to account for both short-range and long-range atomic interactions. Tested on over 150,000 crystal structures, PSCG-Net outperforms the baseline Crystal Graph Convolutional Neural Network model by achieving a mean absolute error of 0.065 eV in formation energy prediction. The model's effectiveness is further supported by consistent results across six diverse data sets and confirmed by first-principles calculations using hybrid functionals in band gap-type predictions. Additionally, PSCG-Net is demonstrated to be practical for screening high-performance materials in photovoltaics, dielectrics, and superconductors. By accurately capturing hierarchical atomic interactions, this approach accelerates the design and discovery of materials and offers a versatile framework applicable to multiscale challenges in various scientific disciplines. This framework not only enhances predictive accuracy but also paves the way for breakthroughs in materials science research and technological innovation.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"100 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145189182","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Constant-pH Simulation of the Human β2 Adrenergic Receptor Inactivation","authors":"Federico Ballabio, Riccardo Capelli","doi":"10.1021/acs.jcim.5c01641","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01641","url":null,"abstract":"Understanding the molecular basis of pH-dependent G protein-coupled receptor (GPCR) signaling is crucial for comprehending physiological regulation and drug design. Here, we investigate the human β₂ adrenergic receptor (β₂AR), a prototypical GPCR whose function is sensitive to pH conditions. Employing extensive constant-pH molecular dynamics simulations, we provide a detailed atomistic characterization of β₂AR inactivation across physiologically relevant pH values (4–9). Our simulations reveal that β₂AR inactivation is closely linked to protonation events at critical residues, notably E268^6×30 involved in the ionic lock formation. Furthermore, we find that inactivation occurs without direct sodium binding to the ion-binding pocket around residue D79^2×50. Instead, sodium ions predominantly interact with D113^3×32, effectively blocking deeper entry toward the traditional binding site. These results challenge existing mechanistic models and highlight the necessity of accurately modeling electrostatics in GPCR simulations. Our findings underscore the potential of constant-pH methodologies to advance the understanding of GPCR dynamics, influencing both fundamental biology and therapeutic strategies.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"31 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145189529","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Activation and Hydroxylation Mechanism of Aromatic C-H and C-F of 3-Fluoro-l-tyrosine Catalyzed by the Heme-Dependent Tyrosine Hydroxylase.","authors":"Yijing Wang,Yongjun Liu","doi":"10.1021/acs.jcim.5c02190","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c02190","url":null,"abstract":"LmbB2 is a peroxygenase-like heme-dependent l-tyrosine hydroxylase (TyrH) that hydroxylates natural l-tyrosine to L-3,4-dihydroxyphenylalanine (DOPA). When challenged with 3-fluoro-l-tyrosine as a substrate, both the C-H and C-F bonds can be hydroxylated, leading to two products, DOPA and 3-F-5-OH-Tyr. However, the crystal structure shows only one binding conformation of the substrate (3-F-Tyr) but two orientations of the fluorine atom, which means that both C-H and C-F are activated. To gain insights into the hydroxylation mechanism, computational models were constructed, and a series of combined QM/MM calculations were performed. Our calculation results reveal that it is the two binding orientations of the substrate that control the final product distribution. Orientations A and B employ different mechanisms for breaking C-H and C-F as well as for hydroxylating the aromatic substrate. Orientation A only corresponds to the C-H hydroxylation, while orientation B is associated with the C-F hydroxylation. The dissociation of the O-O bond in Cpd 0 (Fe(III)-OOH) is in concert with the electron transfer from the iron center to the porphyrin ring, generating the Cpd I intermediate, which is responsible for initiating the reaction. Since the leaving F- takes two electrons away from the substrate, another molecule of hydrogen peroxide is required to complete the catalytic cycle in hydroxylation of C-F bond, and the aromatization of intermediate may occur outside the active site of the enzyme. During the reaction, His88, two crystal water molecules, and the porphyrin ring play critical roles in the proton and electron transfer. Although the hydroxylation of C-H and C-F bonds follows different reaction pathways, they correspond to very similar overall energy barriers; therefore, it is the distribution of two binding orientations of the substrate that determines the final hydroxylated products. These results may provide useful information for understanding the reactions catalyzed by heme-dependent tyrosine hydroxylases.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"43 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145181173","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DeepMaT: Prediction of Target Peptide Classification and Cleavage Site by Combining Mamba2 and Multiple Attention Mechanisms.","authors":"Qianmao Wen,Aoyun Geng,Junlin Xu,Yajie Meng,Leyi Wei,Zilong Zhang,Quan Zou,Feifei Cui","doi":"10.1021/acs.jcim.5c01489","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01489","url":null,"abstract":"Signal peptides and transit peptides are essential for directing mature proteins to their proper cellular locations, particularly through cleavage following transport. Although various prediction tools achieve strong performance in identifying and classifying targeting peptides, their accuracy in determining cleavage sites remains limited. We introduce DeepMaT, a deep learning model that integrates Mamba2 and a multihead self-attention mechanism, leveraging the global modeling capabilities of Mamba2 and the localized focus of self-attention. Experimental results show that DeepMaT significantly outperforms state-of-the-art models in cleavage site prediction, achieving an accuracy of 0.867 for thylakoid transit peptides and also performing well on other peptides. Moreover, DeepMaT can accurately learn the amino acid distribution of real samples. Ablation experiments show that the combination of Mamba and attention mechanisms can improve model efficiency, further proving the effectiveness of the combination. It also enables prediction of targeting peptides with unspecified cleavage sites, offering a valuable tool for protein database annotation. DeepMaT is freely available on GitHub at https://github.com/qianmao2001/DeepMaT.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"52 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145153428","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Freedom Space 3.0: ML-Assisted Selection of Synthetically Accessible Small Molecules.","authors":"Anna Kapeliukha,Serhii Hlotov,Mykola Protopopov,Igor Dzyuba,Maryna Vasylchuk,Dmitriy M Panov,Olga O Tarkhanova,Yurii S Moroz","doi":"10.1021/acs.jcim.5c01912","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01912","url":null,"abstract":"Advances in machine learning (ML) have revolutionized the exploration of chemical space, enabling the creation of subsets tailored for specific applications. Herein, we describe the development of Chemspace Freedom Space 3.0, a chemical library of synthetically accessible small molecules derived from ML-based filtering of building blocks. Our approach employs a model trained on a custom molecular representation to refine the selection of building blocks prior to enumeration, enhancing the quality and synthetic feasibility of the derived molecules. Freedom Space 3.0 comprises 5 billion molecules, generated using ten well-validated chemical transformations, and is complementary to Enamine REAL Space. We computationally evaluate the physicochemical properties, chemical diversity, and synthetic accessibility of the molecules from Freedom Space 3.0. Furthermore, experimental validation demonstrates a success rate of over 80% within a 4-6 week synthesis on a set of 700 molecules, proving the potential for Freedom Space 3.0 to accelerate hit finding and hit follow-up workflows.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"89 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145140510","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Predicting Drug-Drug Interaction via Dual-Drug Visual Representation.","authors":"Lingxuan Xie,Tengfei Ma,Yuqin He,Yiping Liu,Xiangxiang Zeng","doi":"10.1021/acs.jcim.5c01467","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01467","url":null,"abstract":"Drug-drug interaction (DDI) prediction is essential for ensuring medication safety and therapeutic efficacy. While existing models often rely on chemical descriptors or molecular graphs, they tend to overlook the rich spatial and structural cues embedded in visual molecules. To address this issue, we propose DDVR-DDI, a novel vision-based framework that predicts DDIs by encoding drug pairs as a single fused molecular image, enabling direct modeling of their potential interaction interface. To enhance representation learning of visual drug pairs, we introduce a two-stage self-supervised pretraining strategy: a position-invariant contrastive task improves understanding of certain drug pairs in different spatial variations, while a jigsaw puzzle task encourages fine-grained structural understanding. Additionally, we develop a multiexpert voting mechanism, where multiple models analyze distinct augmented views of each drug pair to boost prediction accuracy and stability through ensemble inference. Extensive experiments on benchmark DDI data sets show that our model achieves state-of-the-art performance. To further interpret its predictions, we employ Grad-CAM visualizations and conduct multiple experiments to validate the stability and interpretability of the model; furthermore, we conduct a case study on Ritonavir inhibition of CYP3A, revealing that our model consistently highlights chemically significant substructures. These findings underscore the potential of image-based modeling for both accurate prediction and mechanistic insight in drug interaction research.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"19 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145140470","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Use of Measured Residual Dipolar Couplings to Calculate Residual Dipolar Couplings for a Protein Structure: A Case Study Using Hen Egg-White Lysozyme.","authors":"Maria Pechlaner,Wilfred F van Gunsteren,Niels Hansen,Lorna J Smith","doi":"10.1021/acs.jcim.5c01428","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01428","url":null,"abstract":"Five sets of RDC values for the backbone of [13C,15N]-labeled Hen Egg-White Lysozyme (HEWL, 320 RDCs), obtained from NMR experiments of the protein in an ether bicelle solution at a temperature of 308 K and pH 3.8, were used to calculate RDC values by application of two methods, the alignment-tensor (AT) method and the method of magnetic-field rotational sampling (HRS), applied to five X-ray structures of HEWL, to investigate the relevance of measured RDC values for the structure determination or refinement of proteins. In contrast to other quantities Q observable by NMR, such as NOE intensities or 3J-couplings, for which a relation Q(r) between the quantity Q and a single structure r of a protein can be used to calculate average values ⟨Q(r)⟩, averaged over the Boltzmann-weighted structural ensemble of the protein at finite temperature in solution, an RDC is not defined in terms of a single structure but as an average over a slightly nonuniform rotational and orientation distribution of the protein. This averaging between large positive and negative values reduces the kHz size of a dipolar coupling (DC) by a factor of 103 to 104 to the Hz range of a residual dipolar coupling (RDC). Since the nonuniform orientation distribution can neither be measured nor faithfully mimicked at atomic resolution on a computer, RDC values for a given protein structure are commonly calculated by minimizing the difference between calculated and measured RDC values for a given set of measured target RDC values by varying the orientation distribution of the protein in one way or the other. These three features of RDCs, a very large reduction of size as a result of averaging over orientations, their definition in terms of an unknown, immeasurable orientation distribution, and their calculation using a set of target RDC values, lead to a sensitivity of the calculated RDC values to the size and type of the particular set of RDCs used in the calculation. This reduces the usefulness of measured RDCs for structure determination or refinement of proteins compared to NOE intensities or 3J-couplings.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"30 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145140467","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CSMILES: A Compact, Human-Readable SMILES Extension for Conformations.","authors":"James W Furness,Kevin B Moore,Art Bochevarov","doi":"10.1021/acs.jcim.5c01613","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01613","url":null,"abstract":"While line notation schemes for molecular structure are well developed, they are generally unable to distinguish different conformations of the same molecule. CSMILES, an extension to the ubiquitous line notation scheme, SMILES, has been developed to address this issue. CSMILES are short strings of text that encode information characterizing the conformer structure in the maximally compact form. A conformer is defined by the dihedral angles associated with a structure that has a specified connectivity between atoms. The extension is straightforward: in the simplest case values for the dihedral angles of these bonds are determined from the atomic coordinates and added within a SMILES string at the location of the bond. For example, the canonical SMILES string of pentanol-1 is OCCCCC, and the CSMILES of one of its conformers is O{299}C{180}C{178}C{70}C{56}C. Evidently, the CSMILES strings remain readable, especially for smaller molecules. More difficult cases involving branching, rings, symmetry, and other complications have also been covered by our definitions. Further, CSMILES strings are canonicalized at the conformer level beyond simple connectivity. As such, canonical CSMILES strings are invariant to atom reordering, rigid translation, and rigid rotation. A two-way conversion from three-dimensional (3D) structure to CSMILES has been implemented, and the article is accompanied by a Python code which effectuates such conversions. Possible applications for CSMILES strings are discussed and include efficient storage of 3D structure information as well as development of machine learning models for conformation-dependent properties.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"73 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145153364","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Rethinking Retrosynthesis: Curriculum Learning Reshapes Transformer-Based Small-Molecule Reaction Prediction.","authors":"Rahul Sheshanarayana,Fengqi You","doi":"10.1021/acs.jcim.5c01508","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01508","url":null,"abstract":"Retrosynthesis prediction remains a central challenge in computational chemistry, particularly when models must generalize to rare or structurally complex reactions. We present a curriculum learning (CL) framework that reshapes model training by systematically controlling reaction difficulty during learning, directly addressing the challenge of chemical generalization. In contrast to conventional generative approaches that treat all training reactions uniformly, our method introduces reactions in a chemically informed progression, gradually exposing the model to increasingly complex transformations based on synthetic accessibility, ring complexity, and molecular size. This difficulty-aware pacing allows the model to better capture reaction conditionality, preserve chemical plausibility, and avoid failure modes commonly observed in rare or underrepresented transformations. Applied across three transformer-based architectures─ChemBERTa + DistilGPT2, ReactionT5v2, and BART─the framework yields substantial performance gains. Notably, the largest improvements are observed in the BART model, which lacks any chemical domain pretraining: CL improves its top-1 accuracy from 27.0% to 75.9% (+48.9%). The remainder of our evaluations use ChemBERTa + DistilGPT2 as a representative pretrained model. In low-data regimes with only 50% of the training data, CL increases top-1 accuracy from 16.9% to 46.6% (+29.7%). Under scaffold-based splits, CL improves top-1 accuracy by up to 29%, and in structurally dissimilar settings (Tanimoto similarity <0.4), CL boosts top-1 accuracy from 18.2% to 69.4% (+51.2%), demonstrating strong robustness to distributional shifts. These improvements are achieved without auxiliary labels, templates, or reaction class supervision. Looking forward, this CL framework may aid retrosynthetic route planning for pharmaceutical intermediates, catalysts, polymers, and functional materials.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"91 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145140479","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CNSGT: Generative Transformer for De Novo Drug Design Targeting the Central Nervous System.","authors":"Yingjun Chen,Ding Luo,Shengneng Chen,Tingting Hou,Chao Huang,Weiwei Xue","doi":"10.1021/acs.jcim.5c01541","DOIUrl":"https://doi.org/10.1021/acs.jcim.5c01541","url":null,"abstract":"The design of novel central nervous system (CNS) drugs presents formidable challenges due to the restrictive nature of the blood-brain barrier, which imposes stringent physicochemical requirements. Recent advances in deep learning, particularly Transformer-based architectures, have shown great potential for de novo molecular design. In this study, we present CNSGT, a novel generative framework that integrates variational autoencoders (VAE) with self-attention mechanisms to address the complexity of CNS drug design. By overcoming the limitations of traditional SMILES-based representations, CNSGT effectively captures molecular structure and semantic relationships. The model is pretrained on large-scale molecular data sets and fine-tuned via transfer learning for target-specific generation, demonstrated on dopamine transporter (DAT) inhibitors. The results show that CNSGT generates chemically valid molecules with high CNS drug-likeness (CNS MPO score >4) and improved synthetic accessibility (SAScore <3). The generated molecules also exhibit promising binding affinities in molecular docking (Glide docking score < -8 kcal/mol) and dynamic simulation studies with stable binding conformations. And theoretically prove their good synthetic accessibility through synthetic route analysis by medical chemists, suggesting the model's potential for expanding the useful chemical space and accelerating CNS drug discovery.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"22 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2025-09-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145140466","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}