Computational screening of metal-organic frameworks (MOFs) relies on crystallographic inputs that are commonly treated as “computation-ready”. In practice, however, conventional CIF preprocessing often applies fixed-parameter treatments, overlooking the structural details described in the original reports. To address this, we introduce SmartCIF, a context-aware literature-integrated framework that redefines CIF preprocessing as an explicit assumption-driven procedure. SmartCIF couples topology-based structural analysis with natural-language reasoning over the original publications to make chemically informed decisions about retaining or removing all kind of CIF parts according to the user’s computational objectives. Benchmarking across 321 MOFs against reported BET surface areas and CO2/N2 adsorption data demonstrates that SmartCIF reconciles geometric accessibility with chemical fidelity, avoiding both pore-blocking and over-opened nonphysical results base on the original publications. These results establish that CIF preprocessing is inherently application-dependent and that treating preprocessing assumptions as explicit, controllable variables is essential for reproducible interpretable high-throughput screening. This assumption-aware paradigm embodied by SmartCIF generalizes existing computation-ready resources and provides a flexible foundation for large-scale simulations beyond adsorption.
{"title":"SmartCIF: A Context-Aware Multi-Agent System for Automated Preprocessing and Curation of MOF CIFs","authors":"qixiang zhang, Chen Zhang, Liwei Wang","doi":"10.1039/d6cp00100a","DOIUrl":"https://doi.org/10.1039/d6cp00100a","url":null,"abstract":"Computational screening of metal-organic frameworks (MOFs) relies on crystallographic inputs that are commonly treated as “computation-ready”. In practice, however, conventional CIF preprocessing often applies fixed-parameter treatments, overlooking the structural details described in the original reports. To address this, we introduce SmartCIF, a context-aware literature-integrated framework that redefines CIF preprocessing as an explicit assumption-driven procedure. SmartCIF couples topology-based structural analysis with natural-language reasoning over the original publications to make chemically informed decisions about retaining or removing all kind of CIF parts according to the user’s computational objectives. Benchmarking across 321 MOFs against reported BET surface areas and CO2/N2 adsorption data demonstrates that SmartCIF reconciles geometric accessibility with chemical fidelity, avoiding both pore-blocking and over-opened nonphysical results base on the original publications. These results establish that CIF preprocessing is inherently application-dependent and that treating preprocessing assumptions as explicit, controllable variables is essential for reproducible interpretable high-throughput screening. This assumption-aware paradigm embodied by SmartCIF generalizes existing computation-ready resources and provides a flexible foundation for large-scale simulations beyond adsorption.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"16 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147448431","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Lasse Stausberg,Frank Heberling,Johannes Lützenkirchen
Mineral surfaces in contact with aqueous solutions develop an electric double layer (EDL) through surface (de-)protonation reactions and adsorption of ions, diffusion, and electrostatic forces, resulting in a Stern- and a diffuse layer of ions. Most current models used for surface speciation calculations do not consider changes in surface chemistry caused by charge regulation effects, i.e. effects of interacting EDLs of surfaces in close proximity. Charge regulation modeling requires equilibrium calculation of every involved surface simultaneously, while also solving the Poisson-Boltzmann equation (PBE) to quantify electrostatic interaction. Since analytical solutions of the PBE for complex geometries do not exist it becomes necessary to solve such problems numerically. A Python code is presented that combines a general chemical speciation code, Three Plane Surface Complexation Model, and a Finite Element solution of the PBE on two-dimensional domains. The Finite Element PBE solver is benchmarked against analytical solutions and the speciation code is benchmarked against a PHREEQC model as well as an existing 1D charge regulation code. A test case involving charge regulation in a corner of two perpendicular surfaces is modeled. Charge regulation modeling on a nanoscale enables simulations of the electrostatic environment and surface chemistry in nano-confined systems and interactions of nanoparticles. This may also improve simulations of environmental and biological systems, cementitious materials and modeling of the electrostatic environment and sorption on nanoporous clay materials. Such information can be vital for the in depth understanding of natural and engineered barrier systems of nuclear waste repositories or other environmental scenarios.
{"title":"Charge regulation and surface complexation modeling in nanoscale 2D geometries: benchmarking and test cases of a novel code (CRESCENDO).","authors":"Lasse Stausberg,Frank Heberling,Johannes Lützenkirchen","doi":"10.1039/d6cp00143b","DOIUrl":"https://doi.org/10.1039/d6cp00143b","url":null,"abstract":"Mineral surfaces in contact with aqueous solutions develop an electric double layer (EDL) through surface (de-)protonation reactions and adsorption of ions, diffusion, and electrostatic forces, resulting in a Stern- and a diffuse layer of ions. Most current models used for surface speciation calculations do not consider changes in surface chemistry caused by charge regulation effects, i.e. effects of interacting EDLs of surfaces in close proximity. Charge regulation modeling requires equilibrium calculation of every involved surface simultaneously, while also solving the Poisson-Boltzmann equation (PBE) to quantify electrostatic interaction. Since analytical solutions of the PBE for complex geometries do not exist it becomes necessary to solve such problems numerically. A Python code is presented that combines a general chemical speciation code, Three Plane Surface Complexation Model, and a Finite Element solution of the PBE on two-dimensional domains. The Finite Element PBE solver is benchmarked against analytical solutions and the speciation code is benchmarked against a PHREEQC model as well as an existing 1D charge regulation code. A test case involving charge regulation in a corner of two perpendicular surfaces is modeled. Charge regulation modeling on a nanoscale enables simulations of the electrostatic environment and surface chemistry in nano-confined systems and interactions of nanoparticles. This may also improve simulations of environmental and biological systems, cementitious materials and modeling of the electrostatic environment and sorption on nanoporous clay materials. Such information can be vital for the in depth understanding of natural and engineered barrier systems of nuclear waste repositories or other environmental scenarios.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"8 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147439540","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Nikolaj Klinkby,Anne P Rasmussen,Anders G S Lauridsen,Mordechai Sheves,Lars H Andersen
Retinal protonated Schiff base (RPSB) is the active chromophore in opsin proteins, including rhodopsin for vision. Yet, the spectral consequences of geometric constraints imposed by the protein environment remain insufficiently characterised. We report on gas-phase action-absorption spectra of six retinal analogues with defined steric modifications, recorded in an electrostatic ion-storage ring after cooling in a cryogenic ion trap. Analogues bearing out-of-plane distortions or a shortened π-conjugated polyene chain exhibit pronounced blue-shifts in their absorption maxima. We further present the spectrum of a cryogenically cooled RPSB photofragment of mass 248 amu, whose absorption band near 370 nm matches that of a synthesised β-ionone protonated Schiff base, consistent with substantial truncation of the polyene system. These results isolate the intrinsic spectral signatures of constrained RPSB geometries and provide a framework for understanding protein-induced tuning in opsins.
{"title":"Spectroscopy of cryogenic protonated Schiff-base retinal derivatives.","authors":"Nikolaj Klinkby,Anne P Rasmussen,Anders G S Lauridsen,Mordechai Sheves,Lars H Andersen","doi":"10.1039/d6cp00364h","DOIUrl":"https://doi.org/10.1039/d6cp00364h","url":null,"abstract":"Retinal protonated Schiff base (RPSB) is the active chromophore in opsin proteins, including rhodopsin for vision. Yet, the spectral consequences of geometric constraints imposed by the protein environment remain insufficiently characterised. We report on gas-phase action-absorption spectra of six retinal analogues with defined steric modifications, recorded in an electrostatic ion-storage ring after cooling in a cryogenic ion trap. Analogues bearing out-of-plane distortions or a shortened π-conjugated polyene chain exhibit pronounced blue-shifts in their absorption maxima. We further present the spectrum of a cryogenically cooled RPSB photofragment of mass 248 amu, whose absorption band near 370 nm matches that of a synthesised β-ionone protonated Schiff base, consistent with substantial truncation of the polyene system. These results isolate the intrinsic spectral signatures of constrained RPSB geometries and provide a framework for understanding protein-induced tuning in opsins.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"76 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147439537","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We present a molecular dynamics study revealing that dielectric loss in liquid water in GHz region arises not from isolated molecular rotations, but from collective dipolar correlations spanning more than tens of molecules. By introducing a distance-dependent dipole correlation function, we quantify the spatial extent and temporal evolution of orientational fluctuations contributing to dielectric relaxation. Three distinct peaks identified in the dipole vector correlation at 0.25 nm, 0.53 nm, and 0.75 nm-corresponding to coordinated reorientation among approximately 70 water molecules-indicate a strong link between molecular structure and dielectric behaviour. These findings provide a microscopic basis for understanding dielectric absorption and offer new insights into the design of water-based dielectric systems.
{"title":"Understanding Dielectric Loss in Water via Distance-Dependent Dipole Correlation Functions","authors":"Miki Nakano, Shigenori Tanaka","doi":"10.1039/d5cp03962b","DOIUrl":"https://doi.org/10.1039/d5cp03962b","url":null,"abstract":"We present a molecular dynamics study revealing that dielectric loss in liquid water in GHz region arises not from isolated molecular rotations, but from collective dipolar correlations spanning more than tens of molecules. By introducing a distance-dependent dipole correlation function, we quantify the spatial extent and temporal evolution of orientational fluctuations contributing to dielectric relaxation. Three distinct peaks identified in the dipole vector correlation at 0.25 nm, 0.53 nm, and 0.75 nm-corresponding to coordinated reorientation among approximately 70 water molecules-indicate a strong link between molecular structure and dielectric behaviour. These findings provide a microscopic basis for understanding dielectric absorption and offer new insights into the design of water-based dielectric systems.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"31 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147447853","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
CO 2 capture by amino acids offers promising approach for carbon capture technologies, yet the influence of molecular structure and solvent environment on reaction mechanisms remains least understood. The present study investigates CO 2 capture by glycine, alanine, and serine anions across five environments (gas phase, water, DMSO, glycerol and lactic acid) using density functional theory with implicit solvation. The reaction proceeds via barrierless nucleophilic attack forming a zwitterionic intermediate followed by rate-determining intramolecular proton transfer. Glycerol emerges as the optimal medium exhibiting highly exothermic reaction enthalpies (-50.8 to -53.7 kcal/mol) and stabilized transition states below reactant energy levels, due to extensive hydrogen bonding networks. Structural variations reveal a kineticthermodynamic trade-off in which glycine shows most favorable gas-phase thermodynamics (-21.4 kcal/mol) and lowest barriers (+19.4 kcal/mol), while alanine methyl group introduces steric hindrance and serine hydroxymethyl substituent creates complex solvent-dependent behavior including endothermic reaction in DMSO (+0.4 kcal/mol) from over-stabilization of the serine-DMSO complex. A correlation analysis among the key parameters reveals that CO 2 loading capacity negatively correlates with amino acid hydrogen bond donors (r = -0.59), explaining serine suppressed aqueous activity. Machine learning analysis (Gradient Boosting Regression, R² = 0.85) identifies a molecular weight threshold (~105 g/mol) where side-chain complexity dominates reactivity and demonstrates that solvent hydrogen bond donating capability rather than dielectric constant critically governing capture efficiency. These findings establish glycerol-based formulations with glycine or alanine as superior candidates for industrial CO 2 capture (ΔG 298 = -39 to -43 kcal/mol), highlighting strategic solvent selection for designing tunable amino acid-based carbon capture.
{"title":"Solvent Effects on CO 2 Capture by Simple Amino Acids: An Integrated Density Functional Theory -Machine Learning Approach","authors":"Mukul ., Sandhiya Lakshmanan","doi":"10.1039/d5cp04797h","DOIUrl":"https://doi.org/10.1039/d5cp04797h","url":null,"abstract":"CO 2 capture by amino acids offers promising approach for carbon capture technologies, yet the influence of molecular structure and solvent environment on reaction mechanisms remains least understood. The present study investigates CO 2 capture by glycine, alanine, and serine anions across five environments (gas phase, water, DMSO, glycerol and lactic acid) using density functional theory with implicit solvation. The reaction proceeds via barrierless nucleophilic attack forming a zwitterionic intermediate followed by rate-determining intramolecular proton transfer. Glycerol emerges as the optimal medium exhibiting highly exothermic reaction enthalpies (-50.8 to -53.7 kcal/mol) and stabilized transition states below reactant energy levels, due to extensive hydrogen bonding networks. Structural variations reveal a kineticthermodynamic trade-off in which glycine shows most favorable gas-phase thermodynamics (-21.4 kcal/mol) and lowest barriers (+19.4 kcal/mol), while alanine methyl group introduces steric hindrance and serine hydroxymethyl substituent creates complex solvent-dependent behavior including endothermic reaction in DMSO (+0.4 kcal/mol) from over-stabilization of the serine-DMSO complex. A correlation analysis among the key parameters reveals that CO 2 loading capacity negatively correlates with amino acid hydrogen bond donors (r = -0.59), explaining serine suppressed aqueous activity. Machine learning analysis (Gradient Boosting Regression, R² = 0.85) identifies a molecular weight threshold (~105 g/mol) where side-chain complexity dominates reactivity and demonstrates that solvent hydrogen bond donating capability rather than dielectric constant critically governing capture efficiency. These findings establish glycerol-based formulations with glycine or alanine as superior candidates for industrial CO 2 capture (ΔG 298 = -39 to -43 kcal/mol), highlighting strategic solvent selection for designing tunable amino acid-based carbon capture.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"7 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440409","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Machine learning promises to accelerate material discovery by enabling high-throughput prediction of desirable macro-properties from atomic-level descriptors or structures. However, the limited data available about precise values of these properties has been a barrier, leading to predictive models with limited precision or ability to generalize. This is particularly true of lattice thermal conductivity (LTC): existing datasets of precise (ab initio, DFT-based) computed values are limited to a few dozen materials with little variability. Based on such datasets, we study the impact of transfer learning on both the precision and robustness of a deep learning model (ParAIsite). We start from an existing model (MEGNet 1) and show that significant improvements in predicting high-quality approximations of LTC are obtained through applying transfer learning twice: once on the basis of a pre-training of the model on a large number of materials for a different task (predicting formation energy), and a second time using a medium size dataset (a few thousand materials) of low-quality approximations of LTC (based on the AGL workflow). In other words, greater precision and robustness is obtained after a final training (fine-tuning) of the twice pre-trained model with our high-quality, smaller-scale dataset. We also analyze results obtained from using this higher-precision deep-learning model to scan large numbers of materials from the Material Project Database, in search of low-thermal-conductivity compounds.
{"title":"Two-stage transfer learning for deep learning-based prediction of lattice thermal conductivity.","authors":"Liudmyla Klochko,Mathieu d'Aquin","doi":"10.1039/d5cp04401d","DOIUrl":"https://doi.org/10.1039/d5cp04401d","url":null,"abstract":"Machine learning promises to accelerate material discovery by enabling high-throughput prediction of desirable macro-properties from atomic-level descriptors or structures. However, the limited data available about precise values of these properties has been a barrier, leading to predictive models with limited precision or ability to generalize. This is particularly true of lattice thermal conductivity (LTC): existing datasets of precise (ab initio, DFT-based) computed values are limited to a few dozen materials with little variability. Based on such datasets, we study the impact of transfer learning on both the precision and robustness of a deep learning model (ParAIsite). We start from an existing model (MEGNet 1) and show that significant improvements in predicting high-quality approximations of LTC are obtained through applying transfer learning twice: once on the basis of a pre-training of the model on a large number of materials for a different task (predicting formation energy), and a second time using a medium size dataset (a few thousand materials) of low-quality approximations of LTC (based on the AGL workflow). In other words, greater precision and robustness is obtained after a final training (fine-tuning) of the twice pre-trained model with our high-quality, smaller-scale dataset. We also analyze results obtained from using this higher-precision deep-learning model to scan large numbers of materials from the Material Project Database, in search of low-thermal-conductivity compounds.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"15 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147439535","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The development of new heterogeneous catalysts with well-defined nanostructures has been the focus of chemical industry and academia. Oxide supported metal nanoparticle (NP) often encounters with dynamic structural evolutions under preparation and reaction conditions. Clarifying these structural evolution behaviors of metal NPs is an essential prerequisite for understanding their significant influence on catalytic activities and the rational design of high-performance heterogeneous catalysts. This review aims to delineate the advancements made in the last two decades for identifying the structural evolutions of supported NPs, with a particular focus on establishing the correlation between fundamental energetic descriptors and specific evolution pathways. We discuss how advanced in situ characterization techniques and computational simulations have uncovered the mechanisms by which factors including temperature, NP size, oxide reducibility, and adsorbates govern NP stability. Generally, the thermodynamic instability of NP can give rise to sintering, variations of the metal-oxide interaction can cause encapsulation, and the reactive adsorbates can result in structural fluctuations of NP or single-atom (SA) disintegration. Finally, the challenges and opportunities are proposed for further in-depth investigations on structural evolution issues of oxide supported metal NPs.
{"title":"Structural Evolution Behaviors of Oxide Supported Metal Nanoparticles: A Brief Review","authors":"Houyu Zhu, Dongyuan Liu, Xiaoxin Zhang, Chongchong Wu, Yaoyao Han, Haodong Jiang, Xiaoxiao Gong, Yuhua Chi, Wenyue Guo, Hao Ren","doi":"10.1039/d5cp04437e","DOIUrl":"https://doi.org/10.1039/d5cp04437e","url":null,"abstract":"The development of new heterogeneous catalysts with well-defined nanostructures has been the focus of chemical industry and academia. Oxide supported metal nanoparticle (NP) often encounters with dynamic structural evolutions under preparation and reaction conditions. Clarifying these structural evolution behaviors of metal NPs is an essential prerequisite for understanding their significant influence on catalytic activities and the rational design of high-performance heterogeneous catalysts. This review aims to delineate the advancements made in the last two decades for identifying the structural evolutions of supported NPs, with a particular focus on establishing the correlation between fundamental energetic descriptors and specific evolution pathways. We discuss how advanced in situ characterization techniques and computational simulations have uncovered the mechanisms by which factors including temperature, NP size, oxide reducibility, and adsorbates govern NP stability. Generally, the thermodynamic instability of NP can give rise to sintering, variations of the metal-oxide interaction can cause encapsulation, and the reactive adsorbates can result in structural fluctuations of NP or single-atom (SA) disintegration. Finally, the challenges and opportunities are proposed for further in-depth investigations on structural evolution issues of oxide supported metal NPs.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"39 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440410","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yongna Yuan,Jiahe Kang,Yuanchen Li,Ruisheng Zhang,Wei Su
Molecular representation, as one of the fundamental challenges in artificial intelligence-driven drug discovery, has attracted increasing attention due to its low cost and impressive speed while it is applied in molecular property prediction, drug molecule generation, drug-drug interactions, etc. Numerous models that integrate multi-modal representations have been proposed for molecular representation learning. However, existing methods have not yet considered the IUPAC International Chemical Identifier (InChI) as one of the multi-modal inputs. To address this issue, we propose InChINet, a self-supervised molecular representation learning framework that is pre-trained on 10 million unlabeled molecules. It leverages mutual information across the simplified molecular line input system (SMILES) and InChI. In addition, we present token reordering and token masking for SMILES. Combined with SMILES enumeration, these three strategies introduce domain knowledge and improve the model's stability against syntactic variations in SMILES representations. Benefiting from the introduction of InChI and augmentation strategies, InChINet achieves impressive performance on a wide range of downstream tasks, including molecular property prediction, drug-drug interaction (DDI) prediction, clustering analysis, zero-shot cross-lingual retrieval, and ablation study.
{"title":"InChINet: a self-supervised molecular representation learning framework leveraging SMILES and InChI.","authors":"Yongna Yuan,Jiahe Kang,Yuanchen Li,Ruisheng Zhang,Wei Su","doi":"10.1039/d5cp04869a","DOIUrl":"https://doi.org/10.1039/d5cp04869a","url":null,"abstract":"Molecular representation, as one of the fundamental challenges in artificial intelligence-driven drug discovery, has attracted increasing attention due to its low cost and impressive speed while it is applied in molecular property prediction, drug molecule generation, drug-drug interactions, etc. Numerous models that integrate multi-modal representations have been proposed for molecular representation learning. However, existing methods have not yet considered the IUPAC International Chemical Identifier (InChI) as one of the multi-modal inputs. To address this issue, we propose InChINet, a self-supervised molecular representation learning framework that is pre-trained on 10 million unlabeled molecules. It leverages mutual information across the simplified molecular line input system (SMILES) and InChI. In addition, we present token reordering and token masking for SMILES. Combined with SMILES enumeration, these three strategies introduce domain knowledge and improve the model's stability against syntactic variations in SMILES representations. Benefiting from the introduction of InChI and augmentation strategies, InChINet achieves impressive performance on a wide range of downstream tasks, including molecular property prediction, drug-drug interaction (DDI) prediction, clustering analysis, zero-shot cross-lingual retrieval, and ablation study.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"57 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147439536","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We present a computational framework for calculating the free energy landscapes of host–guest binding using a combination of the on-the-fly probability enhanced sampling (OPES) method and its exploratory variant, OPES-explore. The main advantage of this combined algorithm, referred to as OPESCOM, is its ability to deliver accurate and efficient free energy surfaces using intuitive, suboptimal collective variables that require minimal system-specific optimization. Our algorithm converges the binding affinity estimates within a limited simulation time. It also reproduces the underlying free energy landscapes in quantitative agreement with those generated by much longer OPES simulations that employ sophisticated machine-learned collective variables. Furthermore, the free energy landscapes obtained from the OPESCOM algorithm can identify metastable intermediate states, which can only be distinguished by water coordination descriptors, which are not included in the original set of collective variables used for bias deposition. Thus, it makes the workflow for elucidating host–guest binding mechanisms simple and more scalable without sacrificing accuracy or efficiency. Consequently, our method has the potential to improve computational drug discovery efforts.
{"title":"Free energy landscapes of host–guest binding from adaptive bias enhanced sampling","authors":"Revanth Elangovan, Dhiman Ray","doi":"10.1039/d5cp04649a","DOIUrl":"https://doi.org/10.1039/d5cp04649a","url":null,"abstract":"We present a computational framework for calculating the free energy landscapes of host–guest binding using a combination of the on-the-fly probability enhanced sampling (OPES) method and its exploratory variant, OPES-explore. The main advantage of this combined algorithm, referred to as OPES<small><sub>COM</sub></small>, is its ability to deliver accurate and efficient free energy surfaces using intuitive, suboptimal collective variables that require minimal system-specific optimization. Our algorithm converges the binding affinity estimates within a limited simulation time. It also reproduces the underlying free energy landscapes in quantitative agreement with those generated by much longer OPES simulations that employ sophisticated machine-learned collective variables. Furthermore, the free energy landscapes obtained from the OPES<small><sub>COM</sub></small> algorithm can identify metastable intermediate states, which can only be distinguished by water coordination descriptors, which are not included in the original set of collective variables used for bias deposition. Thus, it makes the workflow for elucidating host–guest binding mechanisms simple and more scalable without sacrificing accuracy or efficiency. Consequently, our method has the potential to improve computational drug discovery efforts.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"51 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440408","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Hong Yan, Dan-Ni Chen, Chen-Xu Hu, Hao Lu, Yao Jie, Yi-Fan Zhang, Jing-Yi Guo
PtxSny intermetallic compounds are widely used catalysts in the propane dehydrogenation reaction(PDH). However, the relationships between their composition, structure and PDH performance remain unclear. In this study, the effects of PtSn catalysts (Pt3Sn, Pt1Sn1, Pt2Sn3) and pure Pt on propane dehydrogenation (PDH) with different Pt/Sn ratios are compared and studied by using density functional theory. The results show that with the increase of Sn content, the electron density of Pt increased, thereby reducing the adsorption of propylene on the catalyst surface, and this electronic effect improves the selectivity of propylene. As the Sn content increases, the generalized coordination number() decreases, and the propylene selectivity increases, while moderate surface roughness(R) aids for the desorption of propylene. By comparing the energy barrier (Pt1Sn1(110)(1.452 eV) > Pt2Sn3(110)(1.248 eV) > Pt3Sn(111)(0.628 eV) > Pt(111)(0.535 eV)) and selectivity parameters (Ediff) (Pt2Sn3(110)(0.965 eV) > Pt1Sn1(110)(0.368 eV) > Pt3Sn(111)(0.116 eV) > Pt(111)(-0.893 eV)), the intermetallic compounds Pt3Sn and Pt2Sn3 can both serve as candidate catalysts for propane dehydrogenation. This is consistent with the experimental results. This work provides a theoretical information for the rational design of high-performance Pt-based intermetallic compound catalysts for the PDH reaction.
{"title":"Theoretical Study on Propane Dehydrogenation Reaction Over PtxSny Intermetallic Compounds","authors":"Hong Yan, Dan-Ni Chen, Chen-Xu Hu, Hao Lu, Yao Jie, Yi-Fan Zhang, Jing-Yi Guo","doi":"10.1039/d6cp00055j","DOIUrl":"https://doi.org/10.1039/d6cp00055j","url":null,"abstract":"PtxSny intermetallic compounds are widely used catalysts in the propane dehydrogenation reaction(PDH). However, the relationships between their composition, structure and PDH performance remain unclear. In this study, the effects of PtSn catalysts (Pt3Sn, Pt1Sn1, Pt2Sn3) and pure Pt on propane dehydrogenation (PDH) with different Pt/Sn ratios are compared and studied by using density functional theory. The results show that with the increase of Sn content, the electron density of Pt increased, thereby reducing the adsorption of propylene on the catalyst surface, and this electronic effect improves the selectivity of propylene. As the Sn content increases, the generalized coordination number() decreases, and the propylene selectivity increases, while moderate surface roughness(R) aids for the desorption of propylene. By comparing the energy barrier (Pt1Sn1(110)(1.452 eV) > Pt2Sn3(110)(1.248 eV) > Pt3Sn(111)(0.628 eV) > Pt(111)(0.535 eV)) and selectivity parameters (Ediff) (Pt2Sn3(110)(0.965 eV) > Pt1Sn1(110)(0.368 eV) > Pt3Sn(111)(0.116 eV) > Pt(111)(-0.893 eV)), the intermetallic compounds Pt3Sn and Pt2Sn3 can both serve as candidate catalysts for propane dehydrogenation. This is consistent with the experimental results. This work provides a theoretical information for the rational design of high-performance Pt-based intermetallic compound catalysts for the PDH reaction.","PeriodicalId":99,"journal":{"name":"Physical Chemistry Chemical Physics","volume":"28 1","pages":""},"PeriodicalIF":3.3,"publicationDate":"2026-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147447852","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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：