Journal of Chemical Information and Modeling 最新文献_第3页

Geometric–Chemical Distance between Protein Surfaces 蛋白质表面之间的几何化学距离

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-09-12 DOI: 10.1021/acs.jcim.6c02785

Himanshu Swami, John M. McBride, Jean-Pierre Eckmann, Tsvi Tlusty

{"title":"Geometric–Chemical Distance between Protein Surfaces","authors":"Himanshu Swami, John M. McBride, Jean-Pierre Eckmann, Tsvi Tlusty","doi":"10.1021/acs.jcim.6c02785","DOIUrl":"10.1021/acs.jcim.6c02785","url":null,"abstract":"Proteinsrecognize, bind, and catalyze through molecular surfaces,where geometry and chemical patterning determine interaction. Comparingthese surfaces requires both a geometric–chemical distanceand a correspondence that relates one complete surface to another.Here, we introduce IFACE (Intrinsic Field–Aligned Coupled Embedding).IFACE derives a symmetric geometric–chemical distance by optimizinga probabilistic coupling over intrinsic geometry, mean curvature,electrostatics, hydrophobicity, and hydrogen-bond propensity. Thesame coupling provides an explicit surface map. For molecular-dynamicsconformers, IFACE distinguishes the same protein from distinct proteinsmore accurately than TM-distance and a Laplace–Beltrami spectraldistance. A Jensen–Shannon distribution distance performs bestin this binary identity test because aggregate surface-feature distributionsalready identify each protein. A distance must also satisfy a globalrequirement: Its pairwise values must place many distinct proteinsurfaces consistently in one space. We therefore tested IFACE acrosssix protein families. It produces the strongest family classificationand clustering among the distributional, spectral, MaSIF, and SurfaceIDcomparisons. The inferred maps preserve geodesic neighborhoods andtransfer heme-centered pocket regions across cytochrome P450 proteins.IFACE therefore provides, from one construction, both a distance betweencomplete protein surfaces and the local map that explains that distance.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"14126–14143"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148973891","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}

引用次数: 0

Benchmarking Protein–Ligand Cofolding Models: Correct Ligand Placement Is Decoupled From Accurate Protein Conformation 对标蛋白质配体共折叠模型：正确的配体位置与准确的蛋白质构象解耦

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-20 DOI: 10.1021/acs.jcim.6c02169

Leon Obendorf, Niklas Piet Doering, Petra Knaus, Gerhard Wolber

{"title":"Benchmarking Protein–Ligand Cofolding Models: Correct Ligand Placement Is Decoupled From Accurate Protein Conformation","authors":"Leon Obendorf, Niklas Piet Doering, Petra Knaus, Gerhard Wolber","doi":"10.1021/acs.jcim.6c02169","DOIUrl":"10.1021/acs.jcim.6c02169","url":null,"abstract":"AI-driven cofolding models have become central toolsfor predictingprotein–ligand complexes, yet it remains unclear whether theresulting structures recover the correct protein conformation or merelyplace the ligand plausibly within an otherwise mispredicted fold.Here, we benchmark four state-of-the-art cofolding models, AlphaFold3,RoseTTAFold3, Boltz-2, and Chai-1, and show that correct ligand placementis systematically decoupled from accurate prediction of the surroundingprotein conformation. As test systems, we use kinases and Class AG protein-coupled receptors (GPCRs), families whose pharmacology dependson well-defined conformational states. To enable a prospective assessment,we evaluate complexes released after the training cutoffs of all testedmodels, together with one earlier nucleotide-bound structure as acontrol. We find that state-defining structural features, includingkinase activation-loop, DFG/DYG geometries, and GPCR intracellularconformations such as ICL3, are frequently mispredicted even whenligand root-mean-square deviation (RMSD) is low. Cofolding adaptivelyremodels the orthosteric pocket around the bound ligand, but thislocal adaptation rarely propagates to the correct global conformation.Incorporating state-annotated templates and filtered multiple sequencealignments (MSAs) improves conformational recovery in some cases buthas little effect in others, and model confidence metrics correlatewith small-molecule placement but cannot report whether the localpocket geometry was recovered. Furthermore, while orthosteric ligandplacement is generally reliable, the single allosteric complex inour set is misplaced by all models, even under explicit guidance.These results identify conformational decoupling as a fundamentallimitation of current cofolding approaches across the systems evaluated,with direct implications for state-selective and structure-based drugdesign.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"14188–14199"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149221703","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}

引用次数: 0

HEIWA: Hybrid Explicit–Implicit Water Attention Affects Force Field Accuracy in Structure-Based Protein Language Model ESM3 混合显式-隐式水注意影响基于结构的蛋白质语言模型ESM3的力场精度

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-23 DOI: 10.1021/acs.jcim.6c01844

Ngai Hei Ernest Ho, Chao Tung Yang, I-Son Ng, Jo-Shu Chang

{"title":"HEIWA: Hybrid Explicit–Implicit Water Attention Affects Force Field Accuracy in Structure-Based Protein Language Model ESM3","authors":"Ngai Hei Ernest Ho, Chao Tung Yang, I-Son Ng, Jo-Shu Chang","doi":"10.1021/acs.jcim.6c01844","DOIUrl":"10.1021/acs.jcim.6c01844","url":null,"abstract":"Protein language models have achieved remarkable successin structureprediction and de novo design. However, their understandingof physical interactions remains limited. For structure-based models,we hypothesize that this limitation stems from the absence of explicitsolvent modeling: water mediates hydrogen bonding, hydrophobic burial,and electrostatic interactions that govern protein thermodynamicsand kinetics, yet current structure-based models treat proteins asif they exist in a vacuum. Using thermophilic protein classificationas a proxy task for force-field awareness, per-residue logit mis-correlatedwith H-bonds as a destabilizing attribute and exhibits significantsolvent-accessible surface area (SASA)-dependent error, despite highoverall classification accuracy. These artifacts point to the water–proteininterface as a source of systematic bias. To address this, we proposeHEIWA (HybridExplicit-Implicit Water Attention), a novel attentionmodule that injects solvent-mediated force-field information intostatic ESM3 residue embeddings via an all-query-to-all-key water–residueattention. Biophysics-guided model selection identifies configurationswhose residue-level outputs are more consistent with the RMSF benchmarkand selected biophysical attributes, while classification performanceremains broadly comparable to the baseline. By forcing the model topredict the protein thermophilicity from the contributions of individualresidues, HEIWA learns biophysical patterns that can be extended toSUPFAM absent from the training data set. HEIWA also largely preservesthe embedding structure of ESM3. Molecular dynamics simulations showimproved kinetic stability under thermophilic conditions, suggestingthe potential to generate structures with greater kinetic stability.To our knowledge, this is the first study to incorporate explicitsolvent encoding into a residue-resolution protein language model.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"13827–13845"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149241655","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}

引用次数: 0

Antimicrobial Peptide Databases and Prediction Tools: Toward a Standard Evaluation Framework 抗菌肽数据库和预测工具：迈向标准评估框架

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-27 DOI: 10.1021/acs.jcim.6c02141

Alejandro Cisterna-García, Álvaro M. González-López, Aspasia Vozi, María A. Esteban, Adrian Egli, Catherine Jutzeler, Jose Palma, Álvaro Sánchez-Ferrer, Juan A. Botía

{"title":"Antimicrobial Peptide Databases and Prediction Tools: Toward a Standard Evaluation Framework","authors":"Alejandro Cisterna-García, Álvaro M. González-López, Aspasia Vozi, María A. Esteban, Adrian Egli, Catherine Jutzeler, Jose Palma, Álvaro Sánchez-Ferrer, Juan A. Botía","doi":"10.1021/acs.jcim.6c02141","DOIUrl":"10.1021/acs.jcim.6c02141","url":null,"abstract":"Antimicrobial resistance(AMR) has a profound impact on animaland human health and is associated with substantial morbidity, mortality,and public health costs. There is a clear need to develop novel, effectiveantibiotic agents, which can overcome the current AMR crisis. Antimicrobialpeptides (AMPs) may offer such a solution and have attracted growingattention for their potential to combat AMR. In parallel, the growingavailability of peptide sequences in public databases has stimulatedthe development of numerous machine learning and deep learning toolsto predict antimicrobial activity computationally. However, it remainsunclear how reliably these tools can be compared, as existing studiesoften rely on heterogeneous data sets and inconsistent evaluationprotocols that may lead to data leakage and inflated performance estimates.This raises a central question: what evaluation criteria and benchmarkresources are needed to enable fair, reproducible, and biologicallymeaningful assessment of AMP prediction tools? We address this questionby focusing specifically on antibacterial peptides (ABPs). We firstprovide an overview of AMP databases relevant to antibacterial activityand compare their content, redundancy, and experimental metadata.We then critically assess existing computational tools for ABP prediction,highlighting key limitations related to data set construction, affinityto certain sequences, data leakage, and inconsistent performance reporting.Based on these limitations, we propose a reference evaluation frameworkdesigned to improve comparability, reproducibility, and practicalutility in ABP prediction. Finally, we provide targeted recommendationsfor AMP databases and future tool development to support more robustprogress in the computational discovery of ABPs.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"13708–13726"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149275052","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}

引用次数: 0

Rule-Defined Multitopology PEG Ligand Library for Virtual Screening 规则定义的虚拟筛选多拓扑PEG配体库

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-29 DOI: 10.1021/acs.jcim.6c02712

Yiting Zhang, Xuhui Wang, Hua Hua, Ping Wei, Junli Li

引用次数: 0

Structural Insights into the Coaggregation Mechanism of TTR and TDP TTR和TDP共聚集机制的结构分析

IF 5.6 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 DOI: 10.1021/acs.jcim.6c02386

Xiaoyu Li,Yuanming Cao,Keqing Zhan,Yanyan Zhu,Qingjie Zhao,Yujie Chen,Huiyu Li

{"title":"Structural Insights into the Coaggregation Mechanism of TTR and TDP","authors":"Xiaoyu Li,Yuanming Cao,Keqing Zhan,Yanyan Zhu,Qingjie Zhao,Yujie Chen,Huiyu Li","doi":"10.1021/acs.jcim.6c02386","DOIUrl":"https://doi.org/10.1021/acs.jcim.6c02386","url":null,"abstract":"The self-aggregation of TTR and TDP proteins is closely associated with the treatment of TDP-43 proteinopathies. Recent experimental evidence suggests that TTR and TDP-43 can coaggregate in the cytoplasm of Amyotrophic lateral sclerosis (ALS) patients and disease models. However, their coaggregation molecular mechanisms are unknown. Herein, we investigated the conformational ensembles of TTR, TDP, and TTR-TDP systems, respectively, employing all-atom replica exchange molecular dynamics (REMD), molecular dynamics (MD), and coarse grain (CG) MD simulation methods. Analyses reveal that TTR-TDP heterohexamer can assemble into closed/open β-barrel structures, but TTR/TDP homohexamer can form bilayer structures, implying the distinct coaggregation property of the TTR-TDP heterohexamer. Compared to TTR/TDP systems, the TTR-TDP system has a higher β-sheet content, longer β-strands and larger β-sheets, suggesting the stronger coaggregation ability of the TTR-TDP heterohexamer. Further analyses reveal that the hydrophobic interactions play the pivotal role in driving the coaggregation of TTR and TDP. These findings may provide a deeper understanding of the pathogenesis of related amyloid diseases and prove the development of novel therapeutic strategies.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"34 1","pages":""},"PeriodicalIF":5.6,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149368322","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}

引用次数: 0

Ga3+ Replacing of Fe3+ as a “Trojan Horse” Antimicrobial Strategy: Modeling the Process within Bacterial Catalase–Peroxidase KatG Ga3+取代Fe3+作为“特洛伊木马”抗菌策略：模拟细菌过氧化氢酶-过氧化物酶KatG的过程

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-16 DOI: 10.1021/acs.jcim.6c02504

Lorenzo Gramolini, Richard López-Corbalán, Jorge García-Millán, Nikoleta Kircheva, Marco Marazzi, Cristina García-Iriepa

{"title":"Ga3+ Replacing of Fe3+ as a “Trojan Horse” Antimicrobial Strategy: Modeling the Process 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":"10.1021/acs.jcim.6c02504","url":null,"abstract":"Replacement of Fe3+ by Ga3+ hasemerged asa promising “Trojan horse” antimicrobial strategy inresponse to the growing threat of antibiotic-resistant bacteria. Becausemany bacterial proteins hijack host iron to sustain essential redoxprocesses, substitution with redox inactive Ga3+, whichclosely resembles Fe3+ in ionic radius but possesses onlyone oxidation state, can disrupt their activity. Among potential targets,bacterial catalase-peroxidase shows particular sensitivity to Fe3+/Ga3+ exchange. However, the structural determinantsand thermodynamic driving forces governing this substitution remainpoorly understood. Here we combine ad hoc force fielddevelopment, molecular dynamics simulations, alchemical free energycalculations, and QM/MM methods to elucidate the molecular basis ofFe3+/Ga3+ exchange in the bacterial catalase-peroxidaseKatG. Our simulations show that the axial coordination environmentof the heme B cofactor is largely preserved upon metal substitution,while the distal ligand dynamically alternates between histidine anda coordinated water molecule. Energy decomposition analysis revealsresidue specific differences in electrostatic and van der Waals interactionsfollowing metal exchange. Free energy calculations indicate that substitutionof iron by gallium in the KatG active site is thermodynamically favorable.The presence of a coordinated water molecule emerges as a key structuralfeature in both metalated states, while distinct binding patternscontribute differently to the overall thermodynamic driving force.These results provide an atomistic description of gallium mediatedmetal substitution in KatG and offer mechanistic insight into gallium-basedantimicrobial strategies, supporting the rational design of new iron-mimeticantibacterial agents.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"14102–14114"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","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}

引用次数: 0

A Multimodal Geometric Learning Framework for Interpretable Peptide Toxicity Prediction 可解释肽毒性预测的多模态几何学习框架

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-19 DOI: 10.1021/acs.jcim.6c02423

Xin Li, Jinzhou Wu, Rui Wang, Donglin He, Yu-Juan Zhang

{"title":"A Multimodal Geometric Learning Framework for Interpretable Peptide Toxicity Prediction","authors":"Xin Li, Jinzhou Wu, Rui Wang, Donglin He, Yu-Juan Zhang","doi":"10.1021/acs.jcim.6c02423","DOIUrl":"10.1021/acs.jcim.6c02423","url":null,"abstract":"Peptide therapeutics are promising drug candidates becauseof theirhigh activity and specificity. However, their potential toxicity posesa major hurdle, making accurate computational toxicity predictionsurgently necessary for early safety screening. Existing methods largelyrely on sequence data or simple graph representations, thereby limitingtheir ability to capture higher-order physicochemical interactions.To address this, we propose MBFGNK, an interpretable multimodal geometriclearning framework, for peptide toxicity prediction. It integratessequence representations from pretrained protein language models,3D structural features via an equivariant graph neural network (EGNN),and physicochemical hypergraph representations capturing global, local,and biologically synergistic residue interactions. A modality attentiongate adaptively fuses these heterogeneous data types, while a Chebyshevpolynomial-based Kolmogorov–Arnold network (ChebyKAN) modelscomplex nonlinear toxicity patterns. We evaluated MBFGNK using 5-foldcross-validation and an independent test set. The experimental resultsshow that the ensemble achieved an AUC of 0.9687, an MCC of 0.8449,an Acc of 0.9221, and an F1 of 0.9205 on the independent test set,suggesting improved performance relative to previously reported baselineresults on the same benchmark. Integrated gradients and modality-levelanalyses further identified prediction-informative motifs, residues,and spatial regions, providing interpretable evidence consistent withknown characteristics of toxic peptides. These results suggest thatMBFGNK is an effective and interpretable framework for peptide toxicityprediction and may assist in downstream safety-oriented peptide screening.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"14068–14088"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149218872","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}

引用次数: 0

Pharmacophore-Conditioned Diffusion Model for Ligand-Based De Novo Drug Design 基于配体的新药物设计的药物团条件扩散模型

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-26 DOI: 10.1021/acs.jcim.6c01046

Amira Alakhdar, Barnabas Poczos, Newell Washburn, Kevin Cusack

{"title":"Pharmacophore-Conditioned Diffusion Model for Ligand-Based De Novo Drug Design","authors":"Amira Alakhdar, Barnabas Poczos, Newell Washburn, Kevin Cusack","doi":"10.1021/acs.jcim.6c01046","DOIUrl":"10.1021/acs.jcim.6c01046","url":null,"abstract":"Developing bioactive molecules remains a central, time-and cost-heavychallenge in drug discovery, particularly for novel targets lackingstructural or functional data. Pharmacophore modeling presents analternative for capturing the key features required for molecularbioactivity against a biological target. In this work, we presentPharmaDiff, a pharmacophore-conditioned diffusion model for 3D moleculargeneration. PharmaDiff employs a transformer-based architecture tointegrate an atom-based representation of the 3D pharmacophore intothe generative process, enabling the precise generation of 3D moleculargraphs that align with predefined pharmacophore hypotheses. Throughcomprehensive testing, PharmaDiff demonstrates superior performancein matching 3D pharmacophore constraints compared to ligand-baseddrug design methods. Additionally, it achieves higher docking scoresacross a range of proteins in structure-based drug design, withoutthe need for target protein structures. By integrating pharmacophoremodeling with 3D generative techniques, PharmaDiff offers a powerfuland flexible framework for rational drug design.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"13768–13782"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149251039","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}

引用次数: 0

Modeling the Soil Dissipation Kinetics of Small Molecules 土壤小分子耗散动力学模拟

IF 6.4 2区化学

Journal of Chemical Information and Modeling Pub Date : 2026-11-09 Epub Date: 2026-10-28 DOI: 10.1021/acs.jcim.6c01812

Vincent-Alexander Scholz, Richard Marchese Robinson, Sevil Payvandi, Steffen Hirte, Timothy J.C. O’Riordan, Johannes Kirchmair

{"title":"Modeling the Soil Dissipation Kinetics of Small Molecules","authors":"Vincent-Alexander Scholz, Richard Marchese Robinson, Sevil Payvandi, Steffen Hirte, Timothy J.C. O’Riordan, Johannes Kirchmair","doi":"10.1021/acs.jcim.6c01812","DOIUrl":"10.1021/acs.jcim.6c01812","url":null,"abstract":"Understanding themetabolic fate of small molecules iscentralto risk assessment and the rational design of plant protection products,pharmaceuticals, cosmetics, and food additives. In environmental systems,dissipation is governed by kinetic processes operating across soil,water, and other compartments. For chemicals released into the environment,especially active ingredients of plant protection products, assessingsoil persistence is an essential regulatory requirement. The abilityto predict soil persistence, ideally in advance of commissioning experiments,is highly valuable for identifying novel active ingredients for plantprotection products. Here, we introduce machine learning models thatpredict the full time course of small-molecule dissipation in soil.These models are explicitly constrained by mechanistic assumptionsencoded as ordinary differential equations (ODEs). By design, thisphysics-constrained parametrization enforces physical plausibility.To the best of our knowledge, our models are the first to predictsoil dissipation kinetics from molecular structures using a constrainedtime course framework. We found that this framework almost alwaysoutperformed a conventional QSAR approach of predicting prefittedkinetic parameters. We further investigated the incorporation of soilmetadata, i.e., predefined and experimentally measured soil parameters,into the model, with mixed results. Our framework supports screeningand design decisions by linking molecular properties and soil metadatato predicted environmental lifetimes and dissipation profiles.","PeriodicalId":44,"journal":{"name":"Journal of Chemical Information and Modeling ","volume":"66 21","pages":"13809–13826"},"PeriodicalIF":6.4,"publicationDate":"2026-11-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149275079","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}

引用次数: 0