Journal of Chemical Theory and Computation最新文献

{"title":"Torsionator: Dihedral Parameterization and Conformer Search Made Easy","authors":"Giorgia Brosio,&nbsp;Alice Triveri,&nbsp;Pietro Vidossich,&nbsp;Sergio Decherchi,&nbsp;Marco De Vivo","doi":"10.1021/acs.jctc.6c01006","DOIUrl":"10.1021/acs.jctc.6c01006","url":null,"abstract":"<p>Neural network potentials (NNPs), trained on quantummechanical(QM) data, can deliver near QM-level accuracy while being much fasterthan QM calculations. In this work, we introduce a new software tool,Torsionator, which currently supports MACE, OBIWAN, and UMA NNPs forenergy minimization of small molecules. Torsionator can analyze aset of selected conformers to parameterize the dihedral angles. Thesedihedral parameters can then be used for molecular mechanics calculationsand simulations. We showcase Torsionator by performing dihedral scanningand parameterization of two representative antiviral compounds: favipiravirdefluoro analog T1105 and emivirine. To thoroughly evaluate its robustnessand general applicability, we benchmarked Torsionator using MACE onthe TorsionNet500 dataset, comprising 500 chemically diverse smallmolecules with reference QM torsional profiles. These results indicateTorsionator as a practical, efficient, and scalable software toolfor dihedral parameterization, enabling its routine integration intomolecular simulation workflows.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"10419–10427"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148933442","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":"Accessing Ice–Water Interfaces with Critical Curvature for Investigating Ice Nucleation at Low Supercooling","authors":"Chuanbiao Zhang,&nbsp;Kai Wu,&nbsp;Yiqun Wang,&nbsp;Mingzhe Shao,&nbsp;Jianjun Wang,&nbsp;Yanting Wang,&nbsp;Ming Li,&nbsp;Xin Zhou","doi":"10.1021/acs.jctc.6c01371","DOIUrl":"10.1021/acs.jctc.6c01371","url":null,"abstract":"<p>Formation of critical nuclei is the essential step offirst-orderphase transitions such as the crystallization of supercooled water.However, probing these nanometer-sized critical nuclei remains extremelychallenging in both experiments and simulations because they formrarely and exist transiently. Here we introduce a general simulationapproach, also amenable to experimental realization, that enablesspatiotemporally controlled formation and long-term stabilizationof critical-sized ice nuclei. By covering a crystalline ice substratewith graphene membranes containing a single nanopore, we show thata spherical-cap ice nucleus forms barrierlessly on the nanopore andcoexists with supercooled water over macroscopic time scales at temperaturesabove a pore-size-dependent threshold but rapidly grows once the temperaturefalls below this threshold. The resulting (meta)stable ice–waterinterface provides direct access to the thermodynamic and kineticproperties of critical ice nuclei, allowing the critical radius, interfacialfree energy, molecular attachment rate, and nucleation rate to bedetermined using standard, unbiased simulations without any additionalrare-event sampling. Beyond simulations, our results suggest a feasibleexperimental strategy for directly probing stabilized critical nucleiand their associated ice–water interfaces at low supercoolingconditions that are otherwise difficult to access. This frameworkovercomes a long-standing barrier in both simulation and experimentalstudies of ice nucleation and is readily extendable to other first-orderphase-transition processes.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"9992–10000"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149010287","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":"Improving All-Atom Molecular Dynamics Models for Quantitative Prediction of Nanopore Blockade Current","authors":"Jingqian Liu,&nbsp;Crystal Rodriguez,&nbsp;Min Chen,&nbsp;Aleksei Aksimentiev","doi":"10.1021/acs.jctc.6c01180","DOIUrl":"10.1021/acs.jctc.6c01180","url":null,"abstract":"<p>All-atom molecular dynamics has become an indispensabletool inthe development of nanopore sensors of biological information. Ina typical nanopore experiment, measurements of ionic current flowingthrough a nanopore report on the chemical structure of biomoleculesthat pass through the nanopore. Alone, such experiments are ofteninsufficient to relate the structure of the biomolecules to the ioniccurrent modulations. The molecular dynamics method can establish sucha relationship directly through a brute force simulation under anapplied electric field. Here, we examine the ability of moleculardynamics force fields to reproduce experimentally measured nanoporeblockade currents produced by single-stranded DNA. Our simulationsshow that none of the “off-the-shelf” force fields (CHARMM36,AMBER Parmbsc1, and DES-AMBER) can reproduce experimental data withthe desired level of accuracy. To improve the accuracy, we examinedand refined interactions between ions, protein nanopores, and DNA,guided by experiments designed specifically for this purpose. Ultimately,the introduction of surgical corrections to nonbonded interactionswithin the CHARMM36 force field produced a favorable agreement betweensimulation and experiment. This refined parameterization, initiallydeveloped for nanopore sensing simulations, may have broader applicationsin computational studies of DNA–protein systems.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"10445–10464"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149049729","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":"Efficient and Transferable Machine Learning Potentials for Zn-Coordinated Zeolitic Imidazolate Frameworks Through Structural Diversity Sampling and Transfer Learning","authors":"Shang-Wei Lin,&nbsp;Yen-Yung Wu,&nbsp;Li-Chiang Lin","doi":"10.1021/acs.jctc.6c01271","DOIUrl":"10.1021/acs.jctc.6c01271","url":null,"abstract":"<p>Zeolitic Imidazolate Frameworks (ZIFs) exhibit remarkablestructuralflexibility that governs many of their functional properties, yetaccurately describing their dynamic behavior remains a significantcomputational challenge. Classical force fields often fail to captureframework flexibility, whereas first-principles methods are prohibitivelyexpensive. In this work, high-fidelity machine learning potentials(MLPs) are first developed for ZIF-8 using the equivariant Allegroarchitecture. To reduce the cost associated with the development ofMLPs, an active-learning-inspired structural selection strategy basedon the smooth overlap of atomic positions (SOAP) is implemented, allowingDFT-level accuracy to be achieved with fewer training configurations.Benchmarking against currently available models demonstrates that,while those models also offer reasonable energy predictions, theirforce errors remain 1 order of magnitude larger than the specializedMLP developed herein with a force mean absolute error of as smallas 0.016 eV/Å. Building upon these findings, a transferable MLPspanning diverse Zn-based ZIFs is further constructed. Moreover, ahierarchical transfer learning workflow is also established to enablerapid adaptation to unseen topologies with improved accuracy usingonly a limited number of additional configurations. As a demonstration,MD simulations powered by the developed MLP successfully capture methane-inducedgate-opening in ZIF-8, resolving the structural transition at subnanosecondresolution. Overall, this work provides a transferable, DFT-accuratemachine learning force field for Zn-based ZIFs.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"10391–10402"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149090872","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":"Can Machine Learning Predict Solvation Effects on Energies and Geometries of Highly Charged Molecules?","authors":"Dario Baum,&nbsp;Lucas Visscher,&nbsp;Amber Pausch","doi":"10.1021/acs.jctc.6c00804","DOIUrl":"10.1021/acs.jctc.6c00804","url":null,"abstract":"<p>Recentmachine-learning models have shown that implicit solvationbased on continuum solvation models can be learned efficiently withgraph neural networks. However, previous studies have focused almostexclusively on neutral or only weakly charged molecules, leaving thestrongly charged regime essentially unexplored. Here, we systematicallyinvestigate which model architectures and architectural componentsare required to predict solvation energies and corresponding forcesfor molecules with total charges ranging from −5 to +5. Usingreference data generated from the conductor-like screening model forrealistic solvation (COSMO-RS) for water, acetonitrile, and cyclohexane,we show that an explicit and robust treatment of molecular chargeis essential for reliable performance across charge states. At thesame time, highly charged systems prove easier to learn than neutraland singly charged ones, whose solvation energies depend more stronglyon subtle structural effects. Finally, we demonstrate that the machine-learningmodels enable stable solvent-aware geometry optimization and can beused for efficient conformational sampling in solution.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"10022–10035"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149090894","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":"DMRG for Multi-Component Quantum Systems: A Symmetry-Aware and Unified Framework","authors":"Xinyu Sun,&nbsp;Haibo Ma","doi":"10.1021/acs.jctc.6c01338","DOIUrl":"10.1021/acs.jctc.6c01338","url":null,"abstract":"<p>Recent advances in high-resolutionspectroscopy and cavityquantumtechnologies have driven growing interest in multicomponent quantumchemistry, where electronic degrees of freedom are treated on an equalfooting with other quantum particles. Here, we present a unified symmetry-awaredensity matrix renormalization group (DMRG) framework for multicomponentquantum chemistry. By decoupling physical model specification fromtensor-network representation and numerical optimization, new multicomponentmodels are introduced through symmetry quantum numbers, local basis,and operators, without modifying the underlying tensor algorithms.Built upon a mature tensor infrastructure, the framework providesa reusable symmetry-aware numerical engine, allowing new models tobe implemented through concise Python-level specifications while fullyreusing the optimized computational backend. The framework is demonstratedthrough symmetry-adapted nuclear-electronic orbital (NEO) and cavityquantum electrodynamics (CQED) DMRG calculations, yielding accurateelectronic structures, excited-state energies, and polaritonic properties.This work establishes a reusable and extensible computational frameworkfor symmetry-aware multicomponent quantum chemistry.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"10150–10164"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149022250","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":"Efficient Tensorized Evaluation of Permutation Invariant Polynomials for Representing Potential Energy Surfaces","authors":"Junhong Li,&nbsp;Kaisheng Song,&nbsp;Hua Guo,&nbsp;Jun Li","doi":"10.1021/acs.jctc.6c01659","DOIUrl":"10.1021/acs.jctc.6c01659","url":null,"abstract":"<p>Permutation invariant polynomials (PIPs), together withrelatedpolynomial-based invariant descriptors such as fundamental invariants(FIs), are widely used in constructing high-fidelity global potentialenergy surfaces (PESs) of molecules that contain identical atoms.Although the monomial symmetrization approach (MSA) enables fast evaluationof PIPs through recursive factorization, it leads to deeply nestedcomputational graphs that may require large memory and are inefficientin modern automatic differentiation frameworks. In this work, we introduceJaxPIP, a JAX-based implementation that reformulates PIP/FI evaluationinto tensorized linear algebra operations. By replacing recursivefactorization with dense matrix operations combined with log-exp transformationand segmented summation, the evaluation becomes regular and GPU-friendly.This allows efficient execution with just-in-time compilation andenables large-scale batch evaluation of energies and forces (as wellas higher-order derivatives). The resulting architecture supportsensemble simulations such as quasi-classical trajectory (QCT), path-integralmolecular dynamics (PIMD), and diffusion Monte Carlo (DMC) in a fullyvectorized manner. As demonstrated in examples, JaxPIP provides apractical route for efficient simulations of molecular systems withscalable GPU execution.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"9940–9953"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149090871","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":"pGM5P-26: A Polarizable Gaussian Multipole Water Model with Transferable Temperature-Dependent Thermodynamics","authors":"Yongxian Wu,&nbsp;Piotr Cieplak,&nbsp;Yong Duan,&nbsp;Ray Luo","doi":"10.1021/acs.jctc.6c01267","DOIUrl":"10.1021/acs.jctc.6c01267","url":null,"abstract":"<p>Water belongs toa small family of liquids whose density maximumlies above the melting point, a signature of their preference foropen, tetrahedral networks. Reproducing this anomalous density–temperaturecurve, together with ambient thermodynamics, remains a demanding testfor water force fields: fixed-charge models trade 298 K accuracy forlarge errors in ρ(<em>T</em>), while point-multipolepolarizable models must exclude the 1–2 and 1–3 intramolecularcontributions that govern molecular anisotropy and, through nonadditivecoupling, shape the long-range field. We present pGM5P-26, a five-pointpolarizable Gaussian Multipole (pGM) water model whose Gaussian-distributedcharges and dipoles remain finite at all distances, giving a unifiedtreatment of short- and long-range electrostatics that retains theseintramolecular contributions in a fully self-consistent induction.Its 14 independent parameters were optimized exclusively at 298 Kagainst the liquid density, oxygen–oxygen radial distributionfunction, and uncorrected finite-box self-diffusion coefficient usingSMAC Bayesian optimization with Hyperband scheduling. Despite thissingle-temperature training, pGM5P-26 gives the lowest absolute densityRMSE and the second-lowest 298 K-anchored density-shape RMSE amongthe ten simulated models over the out-of-training temperatures from245 to 380 K. Its thermal-expansion curve ranks second by both metrics,and its approximate finite-size-corrected self-diffusion curve ranksthird, placing all three temperature dependences among the closestto experiment in the benchmark set. Its temperature of maximum densityis 280.3 K (experiment: 277.1 K), the second closest after TIP4P/2005.The second virial coefficient reaches near-quantitative agreementwith experiment above 773 K. These results show that Gaussian multipolepolarization combined with automated optimization captures water’sdensity anomaly without explicit temperature-dependent training.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"10295–10308"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149091977","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":"Enerzyme: An Efficient Reactive Neural Network Potential Training Framework for Enzyme Cluster Models with Application to Methyltransferases","authors":"Weiliang Luo,Heather J. Kulik","doi":"10.1021/acs.jctc.6c01330","DOIUrl":"https://doi.org/10.1021/acs.jctc.6c01330","url":null,"abstract":"Quantum mechanical (QM) cluster models provide an effective framework for mechanistic studies of enzymatic reactions but QM calculations remain computationally demanding. Neural network potentials (NNPs) offer a promising route to reduce their cost, but enzymes introduce new challenges not faced by NNPs for small molecules, including large system sizes, the need for inclusion of solvent environments, substantial polarization, and charge transfer. Here, we demonstrate an integrated software framework for efficient training of NNPs for mechanistic study of enzymes, with demonstrations on QM-cluster models of S-adenosyl-l-methionine-dependent methyltransferases (MTases). Our Enerzyme code introduces modular, electrostatics-aware NNP architectures and combines automated QM-cluster construction with reactive dataset generation. The Enerzymette subpackage also automates reaction pathway exploration at both NNP and DFT levels of theory. We show that iterative flexible scans and nudged elastic band calculations impose stricter requirements on NNPs than conventionally employed dataset metrics. Nevertheless, we develop an approach to train NNPs on fewer than 1,000 system-specific data points to reproduce reaction energetics and transition-state structures on MTase clusters containing up to 545 atoms, approaching chemical accuracy with respect to the reference. We show that direct supervision of atomic charges and alignment of dielectric screening with QM-cluster calculations substantially improve simulation stability and accuracy and that multitask-learned atomic charges predict correct charge transfer and polarization trends and provide chemically meaningful descriptors of reactivity. Furthermore, we demonstrate transferability across chemically diverse catechol O-methyltransferase substrates, indicating NNPs learn generalizable reactivity patterns as training data is expanded across multiple enzymes. Together, these results establish a foundation for accelerating enzyme mechanistic studies and provide insights for future NNP development for biomolecular reactivity.","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"45 1","pages":""},"PeriodicalIF":5.5,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149152065","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":"Analyzing Initialization Strategies for the Local Unitary Cluster Jastrow Ansatz within the Quantum-Centric Supercomputing Framework","authors":"Grier M. Jones,&nbsp;Maforikan J. Amoussou,&nbsp;Maximilian O. Leach,&nbsp;Hans-Arno Jacobsen","doi":"10.1021/acs.jctc.6c01195","DOIUrl":"10.1021/acs.jctc.6c01195","url":null,"abstract":"<p>Inthis study, we analyze the choice of local unitary cluster Jastrow(LUCJ) ansatz initialization and the sensitivity of the sample-basedquantum diagonalization (SQD) algorithm within the quantum-centricsupercomputing (QCSC) framework. We examine six initialization strategies,including those based on coupled-cluster singles and doubles (CCSD),Møller–Plesset second-order perturbation theory (MP2),data-driven coupled-cluster (DDCC), and trivial (zeroes and random)initializations, across 12 molecular systems and three basis sets(STO-3G, cc-pVDZ, and aug-cc-pVDZ). We find that while the mean absoluteerrors (MAEs) between the alternative and CCSD-initialized <em>t</em>₂-amplitudes span many orders of magnitude,the resulting SQD energies are largely insensitive to this variation.In particular, most initializations recover energies within chemicalaccuracy (±1.6 m<em>E</em>_h) of the CCSD reference,with convergence improving as the basis set size increases. Notably,random initialization achieves performance competitive with CCSD acrossall basis sets, while initialization with zeroes, despite having smallerdeviations from CCSD, yields the worst energy agreement. Our resultshighlight that the proximity to the CCSD initialization is not a reliablepredictor of the quality of electronic energies. We further show thatremoving configuration recovery entirely increases the average energydeviation from the recovery-corrected result to 18.6 m<em>E</em>_h, and that noiseless statevector sampling of the LUCJwave function is frequently outperformed by noisy hardware sampling,indicating that hardware noise assists configuration recovery by broadeningthe sampled determinant space. These findings establish that configurationrecovery within SQD, rather than circuit initialization, is the dominantfactor governing energy accuracy within the active-space range examinedhere (up to 32 electrons in 26 orbitals), and suggest that computationallycheaper initialization strategies are viable alternatives to CCSDfor QCSC workflows at this scale.</p>","PeriodicalId":45,"journal":{"name":"Journal of Chemical Theory and Computation","volume":"22 19","pages":"10078–10090"},"PeriodicalIF":5.8,"publicationDate":"2026-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148973889","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}