Pub Date : 2026-10-16DOI: 10.1021/acs.jcim.6c02504
Lorenzo Gramolini,Richard López-Corbalán,Jorge García-Millán,Nikoleta Kircheva,Marco Marazzi,Cristina García-Iriepa
Abstract Replacement of Fe3+ by Ga3+ has emerged as a promising “Trojan horse” antimicrobial strategy in response to the growing threat of antibiotic-resistant bacteria. Because many bacterial proteins hijack host iron to sustain essential redox processes, substitution with redox inactive Ga3+, which closely resembles Fe3+ in ionic radius but possesses only one oxidation state, can disrupt their activity. Among potential targets, bacterial catalase-peroxidase shows particular sensitivity to Fe3+/Ga3+ exchange. However, the structural determinants and thermodynamic driving forces governing this substitution remain poorly understood. Here we combine ad hoc force field development, molecular dynamics simulations, alchemical free energy calculations, and QM/MM methods to elucidate the molecular basis of Fe3+/Ga3+ exchange in the bacterial catalase-peroxidase KatG. Our simulations show that the axial coordination environment of the heme B cofactor is largely preserved upon metal substitution, while the distal ligand dynamically alternates between histidine and a coordinated water molecule. Energy decomposition analysis reveals residue specific differences in electrostatic and van der Waals interactions following metal exchange. Free energy calculations indicate that substitution of iron by gallium in the KatG active site is thermodynamically favorable. The presence of a coordinated water molecule emerges as a key structural feature in both metalated states, while distinct binding patterns contribute differently to the overall thermodynamic driving force. These results provide an atomistic description of gallium mediated metal substitution in KatG and offer mechanistic insight into gallium-based antimicrobial strategies, supporting the rational design of new iron-mimetic antibacterial agents.
{"title":"Ga3+ Replacing of Fe3+ as a\u0000“Trojan Horse” Antimicrobial Strategy: Modeling the\u0000Process within Bacterial Catalase–Peroxidase KatG","authors":"Lorenzo Gramolini,Richard López-Corbalán,Jorge García-Millán,Nikoleta Kircheva,Marco Marazzi,Cristina García-Iriepa","doi":"10.1021/acs.jcim.6c02504","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02504","url":null,"abstract":"Abstract Replacement of Fe3+ by Ga3+ has emerged as a promising “Trojan horse” antimicrobial strategy in response to the growing threat of antibiotic-resistant bacteria. Because many bacterial proteins hijack host iron to sustain essential redox processes, substitution with redox inactive Ga3+, which closely resembles Fe3+ in ionic radius but possesses only one oxidation state, can disrupt their activity. Among potential targets, bacterial catalase-peroxidase shows particular sensitivity to Fe3+/Ga3+ exchange. However, the structural determinants and thermodynamic driving forces governing this substitution remain poorly understood. Here we combine ad hoc force field development, molecular dynamics simulations, alchemical free energy calculations, and QM/MM methods to elucidate the molecular basis of Fe3+/Ga3+ exchange in the bacterial catalase-peroxidase KatG. Our simulations show that the axial coordination environment of the heme B cofactor is largely preserved upon metal substitution, while the distal ligand dynamically alternates between histidine and a coordinated water molecule. Energy decomposition analysis reveals residue specific differences in electrostatic and van der Waals interactions following metal exchange. Free energy calculations indicate that substitution of iron by gallium in the KatG active site is thermodynamically favorable. The presence of a coordinated water molecule emerges as a key structural feature in both metalated states, while distinct binding patterns contribute differently to the overall thermodynamic driving force. These results provide an atomistic description of gallium mediated metal substitution in KatG and offer mechanistic insight into gallium-based antimicrobial strategies, supporting the rational design of new iron-mimetic antibacterial agents.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"20 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210099","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}
Pub Date : 2026-10-16DOI: 10.1021/acs.jcim.6c02792
Gehan A. Ranepura,Saqibul I. Chowdhury,Elizabeth A. Rosenzweig,Ariën S. Rustenburg,Raquel López-Ríos de Castro,Junjun Mao,John D. Chodera,Sukrit Singh,M. R. Gunner
Abstract The binding affinity of a ligand to a protein is influenced by the protonation and tautomeric states of both partners. However, this relationship remains under-investigated due to the limited availability of computational tools capable of treating charge and tautomer states together in a scalable manner to study clinically relevant systems. Here, we use multi-conformation continuum electrostatics (MCCE) to calculate the protonation and tautomer distributions of nine kinase domains bound to 18 FDA-approved inhibitors across 37 cocrystal structures, while considering their Boltzmann-ensemble. The full pipeline was repeated in triplicate, so that every reported value carries a measured run-to-run uncertainty. Our simulations show that protein net charge and proton distribution remain largely stable even upon binding charged inhibitors. Our results find that individual inhibitor charges are dynamic, frequently increasing or decreasing upon binding a specific protein target. We observe that kinase-inhibitor binding shifts the relative probabilities of low-energy states (ΔG < 2.8 kcal/mol), but does not increase the probability of higher-energy states in the bound form. By considering all low-energy charge states and tautomers, we identify when two solution-state tautomers of nearly equal population resolve upon binding into a single state, creating free energy differences of up to 3.5 kcal/mol between tautomers, and in some complexes differences beyond the resolution of the sampling. In turn, we find that minority species can become the dominant component in the bound state, emphasizing the necessity of considering ensemble-wide protonation and tautomer states to accurately predict protein–ligand binding energetics.
{"title":"Kinase Inhibitors\u0000can Change Protonation or Tautomeric\u0000State upon Binding","authors":"Gehan A. Ranepura,Saqibul I. Chowdhury,Elizabeth A. Rosenzweig,Ariën S. Rustenburg,Raquel López-Ríos de Castro,Junjun Mao,John D. Chodera,Sukrit Singh,M. R. Gunner","doi":"10.1021/acs.jcim.6c02792","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02792","url":null,"abstract":"Abstract The binding affinity of a ligand to a protein is influenced by the protonation and tautomeric states of both partners. However, this relationship remains under-investigated due to the limited availability of computational tools capable of treating charge and tautomer states together in a scalable manner to study clinically relevant systems. Here, we use multi-conformation continuum electrostatics (MCCE) to calculate the protonation and tautomer distributions of nine kinase domains bound to 18 FDA-approved inhibitors across 37 cocrystal structures, while considering their Boltzmann-ensemble. The full pipeline was repeated in triplicate, so that every reported value carries a measured run-to-run uncertainty. Our simulations show that protein net charge and proton distribution remain largely stable even upon binding charged inhibitors. Our results find that individual inhibitor charges are dynamic, frequently increasing or decreasing upon binding a specific protein target. We observe that kinase-inhibitor binding shifts the relative probabilities of low-energy states (ΔG &lt; 2.8 kcal/mol), but does not increase the probability of higher-energy states in the bound form. By considering all low-energy charge states and tautomers, we identify when two solution-state tautomers of nearly equal population resolve upon binding into a single state, creating free energy differences of up to 3.5 kcal/mol between tautomers, and in some complexes differences beyond the resolution of the sampling. In turn, we find that minority species can become the dominant component in the bound state, emphasizing the necessity of considering ensemble-wide protonation and tautomer states to accurately predict protein–ligand binding energetics.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"22 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210137","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}
Pub Date : 2026-10-16DOI: 10.1021/acs.jcim.6c00645
Paula Torren-Peraire,Jonas Verhoeven,Dorota Herman,Hugo Ceulemans,Igor V. Tetko,Jörg K. Wegner
Abstract The identification and annotation of chemotypes is central to structure–activity relationship studies and property optimization in drug discovery. Current computational approaches rely on predefined fragmentation rules or structural similarity to identify these molecular scaffolds, often overlooking the rich information available in reaction data. Here, we introduce CASyM (Chemotype Annotation through Synthesis Mapping), a novel approach that leverages hierarchical reaction data to systematically identify chemotypes within compound libraries. CASyM constructs synthesis graphs from curated reaction data sets, detects major common intermediates, and builds chemotype graphs that capture the hierarchical and interconnected nature of chemical synthesis. SMARTS patterns are extracted to define chemotypes, enabling robust assignment of target molecules, including those absent from the original reaction data. We demonstrate the approach across three case studies using public patent data, proprietary electronic laboratory notebook data from drug discovery projects, and simulated multistep synthesis data for compounds lacking experimental synthesis records. CASyM achieves high coverage of target molecules (85–99%), reveals chemotype-property relationships, and enables chemotype assignment in the absence of experimental data. This reaction-centric approach provides a powerful framework for chemotype annotation, facilitating series exploration and optimization in medicinal chemistry.
摘要化学型的鉴定和注释是药物开发中结构-活性关系研究和性能优化的核心。目前的计算方法依赖于预定义的碎片规则或结构相似性来识别这些分子支架，通常忽略了反应数据中可用的丰富信息。在这里，我们介绍了CASyM (Chemotype Annotation through Synthesis Mapping)，这是一种利用分层反应数据系统地识别化合物文库中的化学型的新方法。CASyM从策划的反应数据集构建合成图，检测主要的常见中间体，并构建捕获化学合成的层次和相互联系性质的化学型图。提取SMARTS模式来定义化学型，从而实现目标分子的稳健分配，包括原始反应数据中缺失的分子。我们通过三个案例研究展示了该方法，这些案例研究使用了公共专利数据、药物发现项目的专有电子实验室笔记本数据以及缺乏实验合成记录的化合物的模拟多步骤合成数据。CASyM实现了靶分子的高覆盖率（85-99%），揭示了化学型-性质关系，并在没有实验数据的情况下实现了化学型分配。这种以反应为中心的方法为化学型标注提供了一个强大的框架，促进了药物化学的系列探索和优化。
{"title":"CASyM: Chemotype\u0000Annotation through Synthesis Mapping","authors":"Paula Torren-Peraire,Jonas Verhoeven,Dorota Herman,Hugo Ceulemans,Igor V. Tetko,Jörg K. Wegner","doi":"10.1021/acs.jcim.6c00645","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c00645","url":null,"abstract":"Abstract The identification and annotation of chemotypes is central to structure–activity relationship studies and property optimization in drug discovery. Current computational approaches rely on predefined fragmentation rules or structural similarity to identify these molecular scaffolds, often overlooking the rich information available in reaction data. Here, we introduce CASyM (Chemotype Annotation through Synthesis Mapping), a novel approach that leverages hierarchical reaction data to systematically identify chemotypes within compound libraries. CASyM constructs synthesis graphs from curated reaction data sets, detects major common intermediates, and builds chemotype graphs that capture the hierarchical and interconnected nature of chemical synthesis. SMARTS patterns are extracted to define chemotypes, enabling robust assignment of target molecules, including those absent from the original reaction data. We demonstrate the approach across three case studies using public patent data, proprietary electronic laboratory notebook data from drug discovery projects, and simulated multistep synthesis data for compounds lacking experimental synthesis records. CASyM achieves high coverage of target molecules (85–99%), reveals chemotype-property relationships, and enables chemotype assignment in the absence of experimental data. This reaction-centric approach provides a powerful framework for chemotype annotation, facilitating series exploration and optimization in medicinal chemistry.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"4 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210103","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}
Abstract Protein kinases regulate signal transduction through phosphorylation, and their dysregulation drives diseases like cancer, autoimmune disorders, and Alzheimer’s, making them key drug targets. Fragment-based drug design offers a powerful route to kinase inhibitors but often neglects kinase-specific structural insights. While KinFragLib, a library consisting of 7486 fragments derived from kinase-ligand complexes, provides a rich resource, exhaustive recombination of its fragments is computationally demanding. Here, we present an efficient exploration strategy that integrates subpocket-guided fragment growing with template docking to navigate this vast chemical space. Our automated pipeline prioritizes kinase-specific structural information, enabling targeted generation of chemically diverse ligands. Applied to the PKA kinase, the approach produced 33758 candidate molecules with promising docking scores, of which 15 were synthesized. Seven molecules reduced kinase activity by more than 50% at an inhibitor concentration of 100 μM, and five of these compounds also showed promising efficacy at a concentration of 10 μM. The IC50 determination ultimately revealed that five of the molecules were effective in the two- to three-digit nanomolar range. Crystallographic validation confirmed the predicted binding mode of the most potent compound. This methodology demonstrates how structure-informed guidance can transform fragment-based design into a scalable route for discovering novel kinase inhibitors.
{"title":"Subpocket-Aware\u0000Fragment-Based Docking Pipeline for\u0000Kinase Ligand Design: Methodology and a PKA Case Study","authors":"Katharina Buchthal,Paula Linh Kramer,Dominik Hubach,Gina Bach,Naemi Wagner,Johannes Krieger,Moritz Klöpper,Michael Daude,Wibke E. Diederich,Andrea Volkamer","doi":"10.1021/acs.jcim.6c01373","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c01373","url":null,"abstract":"Abstract Protein kinases regulate signal transduction through phosphorylation, and their dysregulation drives diseases like cancer, autoimmune disorders, and Alzheimer’s, making them key drug targets. Fragment-based drug design offers a powerful route to kinase inhibitors but often neglects kinase-specific structural insights. While KinFragLib, a library consisting of 7486 fragments derived from kinase-ligand complexes, provides a rich resource, exhaustive recombination of its fragments is computationally demanding. Here, we present an efficient exploration strategy that integrates subpocket-guided fragment growing with template docking to navigate this vast chemical space. Our automated pipeline prioritizes kinase-specific structural information, enabling targeted generation of chemically diverse ligands. Applied to the PKA kinase, the approach produced 33758 candidate molecules with promising docking scores, of which 15 were synthesized. Seven molecules reduced kinase activity by more than 50% at an inhibitor concentration of 100 μM, and five of these compounds also showed promising efficacy at a concentration of 10 μM. The IC50 determination ultimately revealed that five of the molecules were effective in the two- to three-digit nanomolar range. Crystallographic validation confirmed the predicted binding mode of the most potent compound. This methodology demonstrates how structure-informed guidance can transform fragment-based design into a scalable route for discovering novel kinase inhibitors.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"20 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210140","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}
Abstract Natural products (NPs) have been considered a critical source of therapeutic drugs, but prioritizing NP-disease associations remain challenging because supporting evidence is distributed across multiple ingredients, targets, protein interactions, and pathways. Herein, we propose a knowledge graph-assisted machine learning model: Uncertainty-Calibrated Evidence-Chains (UCEC) for prioritizing NP-disease associations. UCEC separates predicted association strength from the evidence-conditioned stability of local mechanistic support. On a proxy benchmark designed to recover graph-supported indirect associations, evidence-chain integration increased Hits@50 from 0.392 to 0.509, with 68.8% of NPs indicating improved per-NP AUPRC. Source-separated validation using HERB 2.0 clinical trial records further supported candidate prioritization. Thus, UCEC provides a reliability-aware approach for prioritizing graph-supported NP-disease candidates and generating traceable mechanistic hypotheses for subsequent validation.
{"title":"UCEC: A Knowledge\u0000Graph-Assisted Machine Learning\u0000for Molecular-Mechanism-Interpretable Natural Product-Disease Association\u0000Prioritization","authors":"Wenyu Lu,Yihui Cai,Xiaoqian Peng,Shulan Mei,Xunkai Yin,Zhe Zheng,Zhenzhen Zhu,Songyun Xu,Hanyuan Zhang,Yue Wang,Weijiang Lin,Xia Zhang,Lihong Hu,Jian Liu","doi":"10.1021/acs.jcim.6c02502","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02502","url":null,"abstract":"Abstract Natural products (NPs) have been considered a critical source of therapeutic drugs, but prioritizing NP-disease associations remain challenging because supporting evidence is distributed across multiple ingredients, targets, protein interactions, and pathways. Herein, we propose a knowledge graph-assisted machine learning model: Uncertainty-Calibrated Evidence-Chains (UCEC) for prioritizing NP-disease associations. UCEC separates predicted association strength from the evidence-conditioned stability of local mechanistic support. On a proxy benchmark designed to recover graph-supported indirect associations, evidence-chain integration increased Hits@50 from 0.392 to 0.509, with 68.8% of NPs indicating improved per-NP AUPRC. Source-separated validation using HERB 2.0 clinical trial records further supported candidate prioritization. Thus, UCEC provides a reliability-aware approach for prioritizing graph-supported NP-disease candidates and generating traceable mechanistic hypotheses for subsequent validation.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"1 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210139","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}
Pub Date : 2026-10-15DOI: 10.1021/acs.jcim.6c02262
Subhasmita Mahapatra,Parimal Kar
Abstract Fibroblast growth factor receptor 3 (FGFR3) kinase plays a crucial role in cellular differentiation, proliferation, and skeletal development. Activation of the kinase domain through phosphorylation regulates the downstream signaling cascades, and its dysregulation leads to human dwarfism, skeletal dysplasia, and multiple cancers. However, the mechanistic basis of FGFR3 regulation by phosphorylation and ligand binding remains unclear. Here, we introduce an integrated GaMD-βVAE-MSM pipeline, a hybrid physics-ML framework for studying kinase activation. Gaussian-accelerated molecular dynamics (GaMD) enhances simulations by reducing the energy barrier. The variational autoencoder with tunable regularization was used for nonlinear dimension reduction and identifying slow collective variables. The Markov state model (MSM) was used for identifying metastable states spanning active-inactive conformations. Our results show that phosphorylation drives pronounced domain rearrangement and activation-loop dynamics. The coordinated movement of the molecular brake and regulatory spine governs the kinase activation. Ligand interaction stabilizes the active phosphorylated state with a free energy of −19 kcal/mol. Protein structure network analysis uncovers the long-range allosteric pathways. The potential regulatory sites involving αCβ4, β34, and β45 loops can be further exploited for allosteric intervention. Overall, this study highlights that phosphorylation and ligand binding dual regulate FGFR3 kinase through allosteric communication networks. The integrated GaMD−βVAE–MSM pipeline provides mechanistic insights into FGFR3 activation and may help the rational design of next-generation FGFR3-targeted therapeutics resistant to gatekeeper mutants.
{"title":"Integrating Gaussian\u0000Accelerated Molecular Dynamics,\u0000β-Variational Autoencoders, and Markov State Models Reveals\u0000the Activation Landscape of FGFR3 Kinase","authors":"Subhasmita Mahapatra,Parimal Kar","doi":"10.1021/acs.jcim.6c02262","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02262","url":null,"abstract":"Abstract Fibroblast growth factor receptor 3 (FGFR3) kinase plays a crucial role in cellular differentiation, proliferation, and skeletal development. Activation of the kinase domain through phosphorylation regulates the downstream signaling cascades, and its dysregulation leads to human dwarfism, skeletal dysplasia, and multiple cancers. However, the mechanistic basis of FGFR3 regulation by phosphorylation and ligand binding remains unclear. Here, we introduce an integrated GaMD-βVAE-MSM pipeline, a hybrid physics-ML framework for studying kinase activation. Gaussian-accelerated molecular dynamics (GaMD) enhances simulations by reducing the energy barrier. The variational autoencoder with tunable regularization was used for nonlinear dimension reduction and identifying slow collective variables. The Markov state model (MSM) was used for identifying metastable states spanning active-inactive conformations. Our results show that phosphorylation drives pronounced domain rearrangement and activation-loop dynamics. The coordinated movement of the molecular brake and regulatory spine governs the kinase activation. Ligand interaction stabilizes the active phosphorylated state with a free energy of −19 kcal/mol. Protein structure network analysis uncovers the long-range allosteric pathways. The potential regulatory sites involving αCβ4, β34, and β45 loops can be further exploited for allosteric intervention. Overall, this study highlights that phosphorylation and ligand binding dual regulate FGFR3 kinase through allosteric communication networks. The integrated GaMD−βVAE–MSM pipeline provides mechanistic insights into FGFR3 activation and may help the rational design of next-generation FGFR3-targeted therapeutics resistant to gatekeeper mutants.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"33 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210141","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}
Pub Date : 2026-10-14DOI: 10.1021/acs.jcim.6c02582
Asma Feriel Khoualdi,Finlay Clark,Joshua T. Horton,Daniel J. Cole
Abstract Prediction of protein–ligand binding modes is a key step in computer-aided drug design, since incorrect pose prediction inevitably limits the achievable accuracy of downstream physics-based affinity predictions. The ASAP-Polaris-OpenADMET antiviral competition was launched in 2025 as a means to prospectively test computational methods across a range of measurements that guide early prioritization during a drug discovery campaign including pose prediction. Here, we describe our two open-source, physics-based entries to the pose prediction challenge, which are based on templated molecular growth with hybrid machine learning/molecular mechanics interatomic potentials. We show that our approach was among the top-performing physics-based methods, and with a retrospective change to the workflow to make greater use of the training data as templates, we obtain a success rate of 74% across SARS-CoV-2 and MERS-CoV main protease target protein structures. Finally, we investigate the importance of receptor side-chain flexibility on retrospective performance. We find that building poses against an ensemble of feasible side-chain conformers is useful in generating improved binding poses but that the tested scoring functions are unable to successfully pick out native structures from the generated ensemble.
{"title":"Evaluation of Physics-Based\u0000Pose Prediction and the\u0000Role of Receptor Flexibility in a Community Antiviral Blind Challenge","authors":"Asma Feriel Khoualdi,Finlay Clark,Joshua T. Horton,Daniel J. Cole","doi":"10.1021/acs.jcim.6c02582","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02582","url":null,"abstract":"Abstract Prediction of protein–ligand binding modes is a key step in computer-aided drug design, since incorrect pose prediction inevitably limits the achievable accuracy of downstream physics-based affinity predictions. The ASAP-Polaris-OpenADMET antiviral competition was launched in 2025 as a means to prospectively test computational methods across a range of measurements that guide early prioritization during a drug discovery campaign including pose prediction. Here, we describe our two open-source, physics-based entries to the pose prediction challenge, which are based on templated molecular growth with hybrid machine learning/molecular mechanics interatomic potentials. We show that our approach was among the top-performing physics-based methods, and with a retrospective change to the workflow to make greater use of the training data as templates, we obtain a success rate of 74% across SARS-CoV-2 and MERS-CoV main protease target protein structures. Finally, we investigate the importance of receptor side-chain flexibility on retrospective performance. We find that building poses against an ensemble of feasible side-chain conformers is useful in generating improved binding poses but that the tested scoring functions are unable to successfully pick out native structures from the generated ensemble.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"68 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210147","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}
Pub Date : 2026-10-14DOI: 10.1021/acs.jcim.6c01714
Adam Bess,Sean Rowland,Chris Alvin,Meisam Adibifard,Supratik Mukhopadhyay
Abstract We present eMolFrag2 and eSynth2, enhanced successors to eMolFrag and eSynth for fragment-based molecular decomposition and reconstruction. eMolFrag2 is a complete reimplementation of eMolFrag in Python, using RDKit and graph-based molecular representations to preserve connectivity, identify free atoms, combine adjacent linkers, and avoid aromaticity ambiguities that impair downstream reconstruction. eSynth2 generalizes fragment recombination by removing hard-coded molecular-property filters, supporting multiple bonds to a single brick atom, and adding a targeted reconstruction mode for soundness validation. Across 36,534 nonredundant DUD-E active compounds, the eMolFrag2 → eSynth2 workflow reconstructed every input molecule exactly (Tanimoto coefficient, TC = 1.0), whereas the original eMolFrag→ eSynth workflow reconstructed 51 of the 100 molecules in a benchmark subset. The upgraded workflow also produced broader recombinant chemical spaces, including higher entropy in a four-molecule diversity analysis (2.78 versus 2.70). Both tools are open-source, platform-agnostic, and available for academic and commercial use.
摘要：我们提出了eMolFrag2和eSynth2，它们是eMolFrag和eSynth的增强继承者，用于基于片段的分子分解和重建。eMolFrag2是在Python中对eMolFrag2的完全重新实现，使用RDKit和基于图的分子表示来保持连通性，识别自由原子，组合相邻的连接物，并避免影响下游重建的芳香性歧义。eSynth2通过移除硬编码的分子属性过滤器，支持单个砖原子的多个键，以及添加用于可靠性验证的目标重构模式来推广片段重组。在36,534个非冗余的ddu - e活性化合物中，eMolFrag2→eSynth2工作流准确地重构了每个输入分子（Tanimoto系数，TC = 1.0），而原始的eMolFrag→eSynth工作流在基准子集中重构了100个分子中的51个。升级后的工作流程还产生了更广泛的重组化学空间，包括四分子多样性分析中的更高熵（2.78 vs 2.70）。这两个工具都是开源的，与平台无关，可用于学术和商业用途。
{"title":"e\u0000 MolFrag\u0000 2\u0000and\u0000 e\u0000 Synth\u0000 2: Sound Fragment-Based\u0000Molecular Decomposition and Reconstruction","authors":"Adam Bess,Sean Rowland,Chris Alvin,Meisam Adibifard,Supratik Mukhopadhyay","doi":"10.1021/acs.jcim.6c01714","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c01714","url":null,"abstract":"Abstract We present eMolFrag2 and eSynth2, enhanced successors to eMolFrag and eSynth for fragment-based molecular decomposition and reconstruction. eMolFrag2 is a complete reimplementation of eMolFrag in Python, using RDKit and graph-based molecular representations to preserve connectivity, identify free atoms, combine adjacent linkers, and avoid aromaticity ambiguities that impair downstream reconstruction. eSynth2 generalizes fragment recombination by removing hard-coded molecular-property filters, supporting multiple bonds to a single brick atom, and adding a targeted reconstruction mode for soundness validation. Across 36,534 nonredundant DUD-E active compounds, the eMolFrag2 → eSynth2 workflow reconstructed every input molecule exactly (Tanimoto coefficient, TC = 1.0), whereas the original eMolFrag→ eSynth workflow reconstructed 51 of the 100 molecules in a benchmark subset. The upgraded workflow also produced broader recombinant chemical spaces, including higher entropy in a four-molecule diversity analysis (2.78 versus 2.70). Both tools are open-source, platform-agnostic, and available for academic and commercial use.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"236 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210143","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}
Pub Date : 2026-10-14DOI: 10.1021/acs.jcim.6c00815
Qixing Huang,Haohua Chang,Liyun Mao,Zihao Shen,Wenhao Tian,Meijuan Zheng,Jun Wang,Honglin Li
Abstract The rapidly growing chemical literature contains invaluable information for pharmaceutical research and molecular biology, yet manual extraction on a large scale is increasingly impractical. Optical chemical structure recognition (OCSR), which converts molecular depictions into machine-readable representations, provides a key route toward automated chemical information extraction. However, existing OCSR methods remain limited in recognizing Markush structures, which are widely used in chemical documents, particularly pharmaceutical patents, because of their structural complexity and the scarcity of annotated data. Here, we introduce CLIP-OCSR, an encoder-decoder framework that integrates a domain-adaptively pretrained CLIP-RN50 vision encoder with a transformer decoder for chemical structure recognition, with a focus on Markush representations. To address data scarcity, we developed a synthetic data generation strategy that produces diverse Markush images paired with pseudo-SMILES representations encoding substituent, frequency, and position variations. Benchmarking demonstrates that CLIP-OCSR achieves competitive performance on standard OCSR benchmarks and shows strong capability in recognizing Markush structures involving substituents and frequency variations. For structures with position variations, CLIP-OCSR generates pseudo-SMILES predictions that retain the corresponding positional information, enabling the downstream enumeration of possible positional isomers. Beyond image-only OCSR, we extend CLIP-OCSR toward multimodal Markush structure extraction by integrating graphical recognition with LLM-based textual variable-definition extraction, enabling downstream structure instantiation. Overall, this work advances automated recognition of complex chemical structures and provides a framework for the structured extraction of Markush information from chemical documents.
{"title":"Bridging the Markush\u0000Gap in Optical Chemical Structure\u0000Recognition via a CLIP-Derived Visual Backbone and Synthetic Data\u0000Generation","authors":"Qixing Huang,Haohua Chang,Liyun Mao,Zihao Shen,Wenhao Tian,Meijuan Zheng,Jun Wang,Honglin Li","doi":"10.1021/acs.jcim.6c00815","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c00815","url":null,"abstract":"Abstract The rapidly growing chemical literature contains invaluable information for pharmaceutical research and molecular biology, yet manual extraction on a large scale is increasingly impractical. Optical chemical structure recognition (OCSR), which converts molecular depictions into machine-readable representations, provides a key route toward automated chemical information extraction. However, existing OCSR methods remain limited in recognizing Markush structures, which are widely used in chemical documents, particularly pharmaceutical patents, because of their structural complexity and the scarcity of annotated data. Here, we introduce CLIP-OCSR, an encoder-decoder framework that integrates a domain-adaptively pretrained CLIP-RN50 vision encoder with a transformer decoder for chemical structure recognition, with a focus on Markush representations. To address data scarcity, we developed a synthetic data generation strategy that produces diverse Markush images paired with pseudo-SMILES representations encoding substituent, frequency, and position variations. Benchmarking demonstrates that CLIP-OCSR achieves competitive performance on standard OCSR benchmarks and shows strong capability in recognizing Markush structures involving substituents and frequency variations. For structures with position variations, CLIP-OCSR generates pseudo-SMILES predictions that retain the corresponding positional information, enabling the downstream enumeration of possible positional isomers. Beyond image-only OCSR, we extend CLIP-OCSR toward multimodal Markush structure extraction by integrating graphical recognition with LLM-based textual variable-definition extraction, enabling downstream structure instantiation. Overall, this work advances automated recognition of complex chemical structures and provides a framework for the structured extraction of Markush information from chemical documents.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"24 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210142","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}
Pub Date : 2026-10-14DOI: 10.1021/acs.jcim.6c02602
Jakub Adamczyk,Adam Staniszewski
Abstract We present scikit-fingerprints, a comprehensive, fully scikit-learn-compatible library for molecular machine learning in Python, based on RDKit. Molecular fingerprints and related functionalities are workhorses of chemoinformatics, yet the widely used open-source frameworks are not compatible with the wider Python machine-learning ecosystem based on scikit-learn conventions. scikit-fingerprints closes this gap, bringing molecular fingerprints, molecular filters, similarity and distance measures, applicability domain estimation, data-splitting strategies, and more under a single familiar interface. Scikit-learn compatibility means that an entire chemoinformatics workflow, from a raw SMILES string to a deployable model, can be assembled from composable building blocks and can reuse the mature tooling of the surrounding ecosystem. The underlying RDKit code makes the library familiar and extensible for custom chemoinformatics use cases. We put a strong focus on unified interfaces, ease of use, computational efficiency, customization, and extensibility. scikit-fingerprints makes molecular machine learning faster to prototype, easier to reproduce, and simpler to deploy.
{"title":"scikit-fingerprints\u0000 : A Python Library\u0000for Scikit-Learn-Compatible Molecular Fingerprints and Chemoinformatics","authors":"Jakub Adamczyk,Adam Staniszewski","doi":"10.1021/acs.jcim.6c02602","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02602","url":null,"abstract":"Abstract We present scikit-fingerprints, a comprehensive, fully scikit-learn-compatible library for molecular machine learning in Python, based on RDKit. Molecular fingerprints and related functionalities are workhorses of chemoinformatics, yet the widely used open-source frameworks are not compatible with the wider Python machine-learning ecosystem based on scikit-learn conventions. scikit-fingerprints closes this gap, bringing molecular fingerprints, molecular filters, similarity and distance measures, applicability domain estimation, data-splitting strategies, and more under a single familiar interface. Scikit-learn compatibility means that an entire chemoinformatics workflow, from a raw SMILES string to a deployable model, can be assembled from composable building blocks and can reuse the mature tooling of the surrounding ecosystem. The underlying RDKit code makes the library familiar and extensible for custom chemoinformatics use cases. We put a strong focus on unified interfaces, ease of use, computational efficiency, customization, and extensibility. scikit-fingerprints makes molecular machine learning faster to prototype, easier to reproduce, and simpler to deploy.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"63 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149210144","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：