Journal of Chemical Theory and Computation最新文献

{"title":"Autopylot: Pragmatic Benchmarking of Excited-State Electronic Structure.","authors":"Gregory M Curtin,Madeline L Thomas,Elisa Pieri","doi":"10.1021/acs.jctc.5c00734","DOIUrl":"https://doi.org/10.1021/acs.jctc.5c00734","url":null,"abstract":"Accurate excited-state modeling in photochemical studies hinges on the choice of the electronic structure method, which governs predicted pathways and mechanistic reliability. Yet this selection remains a major challenge, typically relying on chemical intuition and manual screening at a single geometry while overlooking broader regions of the potential energy surface. To overcome these limitations, we developed Autopylot, a Python package that automates excited-state benchmarking by comparing single-structure absorption spectra against a reference across multiple geometries, targeting accurate descriptions of both the Franck-Condon region and excited-state minima. Designed for flexibility, Autopylot supports the seamless addition of new geometry types and electronic structure methods. Reflecting a pragmatic philosophy, it incorporates computational time as a metric, guiding users toward optimal cost-accuracy trade-offs upon request. We benchmarked Autopylot on a set of 28 small organic molecules, where it consistently identified methods that closely reproduce the reference spectra within minutes. This performance marks a major step toward the high-throughput, automated selection of excited-state electronic structure methods.","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"699 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2025-07-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144613073","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Upper-Order TICA and Fractional Non-Markovian Process to Model Anomalous Dynamic Regimes.","authors":"Arnaldo Rapallo","doi":"10.1021/acs.jctc.5c00540","DOIUrl":"https://doi.org/10.1021/acs.jctc.5c00540","url":null,"abstract":"The coupling of time-lagged independent component analysis (TICA) with the Markov state model (MSM) technique has become a well-established route to study dynamics in complex molecular systems. Identification of the slow modes relevant to the molecular functions, quantification of the characteristic times involved in the slow dynamics, and prediction of dynamic properties are the basic frame of application of such methods. Among the current research developments in the field, great activity is devoted to the formulation of methods to improve approximation of the leading eigenfunctions of the transfer operator of a dynamical system from trajectory data and to include memory effects into MSM analysis. Along these lines of research, various developments are proposed here, in the framework of TICA-MSM approaches: a criterion to select dynamically informative intramolecular distances and a method to use them to build optimal nonlinear basis sets for TICA (upper-order TICA) are presented to overcome the limitations of linear approximations to the transfer operator eigenfunctions. Then, a fractional, non-Markovian process is introduced to deal with anomalous dynamic regimes characterized by nonexponential relaxations. The fractional process is described in terms of a time derivative of noninteger order α > 0 in the master equation of the temporal evolution of the states' probabilities, which replaces the exponential decay in time, typical of Markovian processes, with Mittag-Leffler functions in the temporal variable. This kind of temporal dependency is more appropriate to capture the characteristics of anomalous dynamics, often observed in proteins and peptides by both experiments and simulations. The theory is cast in a form that the researchers are familiar with when applying MSM analysis, allowing direct manipulations over the transition probability matrix. Moreover, the technique allows us to check whether the dynamics encoded in the molecular dynamics (MD) trajectory occur in an anomalous regime or not, and, in case, permits to quantify and treat the anomaly by identifying the appropriate fractional order α of the non-Markovian process. Purely Markovian dynamic regimes are special cases of the proposed theory and can be recovered by letting α = 1. The benchmark MD trajectories of chignolin (a), villin (b), and Trp-cage (c) proteins, provided by D.E. Shaw Research (DESRES), are revisited in light of the proposed developments, and cases (a) and (c) show that the ability to describe the system dynamics in terms of fractional non-Markovian processes is necessary to obtain a more accurate qualitative and quantitative picture of molecular dynamics occurring in anomalous regimes.","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"191 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2025-07-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144604077","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Discovering Molecular Insights in Organic Optoelectronics with Knowledge-Informed Interpretable Deep Learning.","authors":"Qian Zhang,Hengyue Zhang,Zhiyao Su,Yajing Sun,Wenping Hu","doi":"10.1021/acs.jctc.5c00713","DOIUrl":"https://doi.org/10.1021/acs.jctc.5c00713","url":null,"abstract":"Deep learning holds significant promise for accelerating molecular screening and materials design. However, the black-box nature of current models limits their ability to generate fundamentally new chemical knowledge and insights. Here, we propose LUMIA (Learning and Understanding Molecular Insights with Artificial Intelligence), an innovative interpretable deep learning framework integrating chemistry-informed contrastive learning and Monte Carlo tree search (MCTS). LUMIA is pretrained on approximately 1.4 million organic molecules, using knowledge-informed augmentations that embed π-conjugation and substituent effects explicitly. This allows it to effectively capture hierarchical molecular representations aligned with chemical intuition. Critically, the explicit integration of chemical knowledge enables LUMIA to achieve state-of-the-art performance across multiple organic optoelectronic property prediction tasks. Leveraging its intrinsic interpretability through MCTS, LUMIA directly uncovers previously unexplored substructure patterns influencing reorganization energy, enabling rational molecular design beyond the training data set. Furthermore, LUMIA reveals novel chemical insights, including synergistic effects of substituent positions in pyrazole derivatives. This study highlights the pivotal role of knowledge embedding in interpretable deep learning, transforming molecular design, and accelerating chemical discovery.","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"21 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144594395","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Zero Excitation Energy Theorem and the Spin-Flip Kernel.","authors":"Tai Wang,Hao Li,Yi Qin Gao,Yunlong Xiao","doi":"10.1021/acs.jctc.5c00714","DOIUrl":"https://doi.org/10.1021/acs.jctc.5c00714","url":null,"abstract":"This work establishes the zero excitation energy theorem, which ensures that the TDDFT equations constructed from an open-shell reference state must admit excited-state solutions with zero excitation energy. This theorem holds exactly in TDDFT but only approximately when the Tamm-Dancoff approximation is used. From this theorem, we derive an identity connecting the spin-conserving and spin-flip kernels. Based on this identity, a method to construct the spin-flip kernel solely from the spin-conserving kernel is proposed. This method is applicable to all types of collinear functional, is numerically stable, and preserves the expected energy degeneracy. Since this spin-flip kernel is merely a simple geometric average of the spin-conserving kernel, the spin-flip TDDFT based on it is easy to implement, especially in programs that already support spin-conserving TDDFT. Numerical tests show that the spin-flip TDDFT and its analytic gradient are as efficient as spin-conserving TDDFT, making them practical for routine use.","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"695 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144603884","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An Improved Virtual Orbital Driven Similarity Renormalization Group Approach for Core-Ionized and Core-Excited States.","authors":"Meng Huang,Francesco A Evangelista","doi":"10.1021/acs.jctc.5c00457","DOIUrl":"https://doi.org/10.1021/acs.jctc.5c00457","url":null,"abstract":"This work combines the multireference driven similarity renormalization group (DSRG) with a reference state obtained using improved virtual orbitals (IVOs) and generalized active space configuration interaction (GASCI) to model core-ionized and core-excited states without costly orbital optimizations. We test the accuracy of the resulting IVO-GASCI-DSRG method combined with three truncation levels across four data sets of molecules containing first-row elements (small molecules, potential energy surfaces, small-to-medium molecules, and X-ray absorption spectra). It is found that the IVO-GASCI-DSRG approach with an active space consisting of three GAS spaces and third-order perturbative corrections (IVO-GASCI[3]-DSRG-MRPT3) strikes the best balance between cost and accuracy. This method exhibits good agreement with the most accurate DSRG truncation scheme based on self-consistent orbitals on small-molecule benchmarks, and it is capable of accurately predicting the potential energy surfaces of core-excited and core-ionized states of CO, N2, and HF. To demonstrate the applicability of this method to medium-sized molecules, we simulate the X-ray absorption spectra of thymine and adenine using IVO-GASCI-DSRG-MRPT3, successfully reproducing key experimental spectral features.","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"29 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2025-07-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144604080","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fusing Domain Knowledge with a Fine-Tuned Large Language Model for Enhanced Molecular Property Prediction.","authors":"Liangxu Xie,Yingdi Jin,Lei Xu,Shan Chang,Xiaojun Xu","doi":"10.1021/acs.jctc.5c00605","DOIUrl":"https://doi.org/10.1021/acs.jctc.5c00605","url":null,"abstract":"Although large language models (LLMs) have flourished in various scientific applications, their applications in the specific task of molecular property prediction have not reached a satisfactory level, even for the specific chemistry LLMs. This work addresses a highly crucial and significant challenge existing in the field of drug discovery: accurately predicting the molecular properties by effectively leveraging LLMs enhanced with profound domain knowledge. We propose a Knowledge-Fused Large Language Model for dual-Modality (KFLM2) learning for molecular property prediction. The aim is to utilize the capabilities of advanced LLMs, strengthened with specialized knowledge in the field of drug discovery. We identified DeepSeek-R1-Distill-Qwen-1.5B as the optimal base model from three DeepSeek-R1 distilled LLMs and one chemistry LLM named ChemDFM, by fine-tuning with the ZINC and ChEMBL datasets. We obtained the SMILES embeddings from the fine-tuned model and subsequently integrated the embeddings with the molecular graph to leverage complementary information for predicting molecular properties. Finally, we trained the hybrid neural network on the combined dual modality inputs and predicted the molecular properties. Through benchmarking on regression and classification tasks, our proposed method can obtain higher prediction performance for nine out of ten datasets in the downstream regression and classification tasks. Visualization of the output of hidden layers indicates that the combination of the embedding with the molecular graph can offer complementary information to further improve the prediction accuracy compared with either the LLM embedding or the molecular graph inputs. Larger models do not inherently guarantee superior performance; instead, their effectiveness hinges on our ability to leverage relevant knowledge from both pretraining and fine-tuning. Implementing LLMs with domain knowledge would be a rational approach to making precise predictions that could potentially revolutionize the process of drug development and discovery.","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"4 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2025-07-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144586633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Localized Active Space State Interaction Singles.","authors":"Matthew R Hermes, Bhavnesh Jangid, Valay Agarawal, Laura Gagliardi","doi":"10.1021/acs.jctc.5c00387","DOIUrl":"10.1021/acs.jctc.5c00387","url":null,"abstract":"<p><p>We introduce localized active space state interaction singles (LASSIS), a multireference electronic structure method that uses two-step diagonalization to model low-energy electronic states of systems characterized by multiple distinct localized centers of strong electron correlation, with weaker but not negligible electron correlation between the centers. LASSIS is a specific variant of localized active space state interaction (LASSI), which restores interfragment interactions omitted by a LASSCF reference wave function by expanding the interacting wave function in a basis of model states characterized by various charge and spin distributions. These distributions, and the number of states of each type, are determined automatically, without any user input, in contrast to previous work with the LASSI formalism. LASSIS combined with multiconfiguration pair-density functional theory (MC-PDFT) energy calculation is shown in test calculations to qualitatively reproduce the results of converged DMRG-PDFT calculations on multimetallic transition metal compounds.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":" ","pages":"6446-6463"},"PeriodicalIF":5.7,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144300633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Neural Mulliken Analysis: Molecular Graphs from Density Matrices for QSPR on Raw Quantum-Chemical Data.","authors":"Oleg I Gromov","doi":"10.1021/acs.jctc.5c00425","DOIUrl":"10.1021/acs.jctc.5c00425","url":null,"abstract":"<p><p>Here, molecular graphs derived from the one-electron density matrix are introduced within a more general effort to explore whether incorporating electronic structure awareness allows a single model to both better generalize from small data and better learn molecular encodings. Diagonal density matrix blocks serve as atomic node embeddings, while off-diagonal blocks provide embeddings for <i>\"link\"</i> nodes related to atomic pairs. In a minimal basis, these embeddings have dimensions of only 45 and 81, yet no information is lost and the original density matrix can be fully reconstructed. Blocks from the basis set overlap matrix are used as edge embeddings to encode structural information and as weights for message aggregation operations. Additionally, element-wise multiplication performed during aggregation may provide access to electronic charges, analogous to Mulliken population analysis. The proposed concept was evaluated using data from the First and Second Solubility Challenges (Llinàs et al. <i>J.Chem. Inf. Model.</i> <b>2008</b>, <i>48</i>, 1289-1303; Llinàs and Avdeef <i>J. Chem. Inf. Model.</i> <b>2019</b>, <i>59</i>, 3036-3040). A graph neural network (GNN) trained on sets of 94 and 1000 drug-like molecules achieved improved solubility prediction accuracy (RMSE 0.63, <i>R</i>² 0.79 in SC-1 and RMSE of 0.83 and 0.92, <i>R</i>² of 0.57 and 0.79 on the \"tight\" and \"loose\" SC-2 test sets, respectively). If combined with existing techniques for predicting electron density from molecular structures, this approach is promising for addressing a range of chemical machine-learning problems.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":" ","pages":"6380-6393"},"PeriodicalIF":5.7,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144504165","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"","authors":"Qiyuan Zhao,&nbsp;Veerupaksh Singla,&nbsp;Hsuan-Hao Hsu and Brett M. Savoie*,&nbsp;","doi":"","DOIUrl":"","url":null,"abstract":"","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"21 13","pages":"XXX-XXX XXX-XXX"},"PeriodicalIF":5.7,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.acs.org/doi/pdf/10.1021/acs.jctc.5c00279","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144569069","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"","authors":"Jiaji Zhang,&nbsp;Jian Liu* and Lipeng Chen*,&nbsp;","doi":"","DOIUrl":"","url":null,"abstract":"","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"21 13","pages":"XXX-XXX XXX-XXX"},"PeriodicalIF":5.7,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://pubs.acs.org/doi/pdf/10.1021/acs.jctc.5c00224","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144569087","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"化学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}