[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85271-en":3,"doc-seo-85271-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},85271,1374391974585,"Genevieve","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","Transferable Implicit Solvent Machine Learning Potential for Drugs and Proteins Approaching Ab Initio Accuracy","Machine learning interatomic potentials (MLPs) transform atomistic modeling but remain limited by slow inference compared with classical force fields, restricting use for biomolecular timescales extending to microseconds and beyond. Implicit-solvent MLPs can mitigate this gap, yet coarse-grained data constraints have historically reduced accuracy. The work presents TWIN, a transferable implicit-water model parameterized by an equivariant graph neural network and trained only on ab initio and experimental labels, delivering strong crystallographic and NMR benchmark performance.","arXiv :2607 . 10887v1 [physics .chem-ph] 12 Jul 2026  \nTransferable Implicit Solvent Machine Learning Potential for Drugs and Proteins Approaching Ab Initio Accuracy  \nJan Eckwert 1 and Julija Zavadlav 1,2*  \n1 Multiscale Modeling of Fluid Materials, Department of Engineering Physics and Computation,  \nTUM School of Engineering and Design and Department of Physics, TUM School of Natural Sciences, Technical University of Munich, Munich, 80333, Germany.  \n2 Atomistic Modeling Center, Munich Data Science Institute, Technical University of Munich, Garching, 85748, Germany.  \n*Corresponding author(s). E-mail(s): [julija.zavadlav@tum.de](julija.zavadlav@tum.de) ;  \nAbstract  \nMachine learning interatomic potentials (MLPs) have revolutionized atomistic modeling, offering the potential to replace traditional methods like Density Functional Theory (DFT) . However, inference time of MLPs is orders of magnitude slower than that of classical force fields, hindering real-world applications for biomolecular systems that require timescales of microseconds and beyond. Implicit solvent MLPs can address this issue, but are faced with data challenges associated with coarse-grained modeling. Consequently, previous approaches relied on empirical force field data, thereby inherently limiting the MLP’s accuracy. Here, we introduce the Transferable Water Implicit Network (TWIN), an implicit water MLP parametrized entirely by an Equivariant Graph Neural Network and trained solely on ab initio and experimental labels. We demonstrate TWIN’s transferability across drug-like molecules, peptides, and proteins, achieving excellent results on ab initio and experimental crystallographic and NMR benchmarks, consistently outperforming previous machine-learning-based implicit solvent or coarse-grained models. Furthermore, TWIN closely matches DFT-based explicit solvent MLPs while providing a two-order-of-magnitude faster timestep evaluation, paving the way for efficient ab initio-level modeling of biomolecular systems in aqueous environments.  \nIntroduction  \nUnderstanding protein dynamics is essential for elucidating the molecular mechanisms underlying biological function. Functional processes, such as transitions between metastable conformational states, allosteric regulation, and protein-drug interactions, are governed by the exploration of complex conformational ensembles [1–3] . Accurately predicting these ensembles and their interactions with drug-like molecules is therefore critical for modern medicine and molecular-level understanding of life [4, 5] . The atomistic description of biologically relevant protein systems, which often encompass tens to hundreds of thousands of atoms, remains computationally prohibitive for ab initio electronic-structure methods such as density functional theory (DFT) . As a result, all atom molecular dynamics (MD) simulations employing classical empirical force fields are the main workhorse for modeling protein conformations, conformational transitions, and protein-ligand interactions [6–8] . While these models provide the computational efficiency necessary to treat large biophysical systems, their reliance on empirical parameterizations imposes inherent limitations on their predictive capabilities [9, 10] .  \nThe emergence of Machine Learning Potentials (MLPs) has fundamentally transformed atomistic modeling, enabling simulations of million-atom systems with accuracies approaching those of quantum-mechanical methods [11, 12] . Symmetry-equivariant message passing graph neural network architectures such as MACE [13] have been central to this progress, as they provide complex representations of molecular potential-energy surfaces while  \nrigorously respecting the underlying symmetries of atomic systems. In parallel, the availability of large, diverse quantum-chemical datasets has substantially expanded the applicability of these models across broad regions of chemical space [14, 15] . A prominent example is the SPICE dataset [","cbCailklMJ0b0hAj","https://ap.wps.com/l/cbCailklMJ0b0hAj","pdf",9217090,3,1,23,"English","en",105,"# Abstract\n# Introduction\n## Protein dynamics and conformational ensembles\n## Limits of classical force fields and ab initio methods\n## Machine learning potentials and equivariant GNN architectures\n## Implicit solvent and coarse-grained approaches","[{\"question\":\"What problem prevents conventional MLPs from being widely used for biomolecular simulations?\",\"answer\":\"MLP inference time is orders of magnitude slower than classical force fields, making it difficult to reach microsecond-scale dynamics required for many biomolecular processes.\"},{\"question\":\"How does TWIN improve implicit-solvent modeling accuracy?\",\"answer\":\"TWIN is an implicit-water MLP trained entirely on ab initio and experimental labels, avoiding reliance on empirical force-field data that previously limited accuracy.\"},{\"question\":\"What performance benefits does TWIN provide compared with existing approaches?\",\"answer\":\"TWIN transfers well across drug-like molecules, peptides, and proteins, matches DFT-based explicit-solvent MLP quality closely, and enables about a two-order-of-magnitude faster timestep evaluation.\"}]",1784202193,58,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"transferable-implicit-solvent-machine-learning-potential-for-drugs-and-proteins-approaching-ab-initio-accuracy","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/transferable-implicit-solvent-machine-learning-potential-for-drugs-and-proteins-approaching-ab-initio-accuracy/85271/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-24","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem prevents conventional MLPs from being widely used for biomolecular simulations?","Question",{"text":75,"@type":76},"MLP inference time is orders of magnitude slower than classical force fields, making it difficult to reach microsecond-scale dynamics required for many biomolecular processes.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does TWIN improve implicit-solvent modeling accuracy?",{"text":80,"@type":76},"TWIN is an implicit-water MLP trained entirely on ab initio and experimental labels, avoiding reliance on empirical force-field data that previously limited accuracy.",{"name":82,"@type":73,"acceptedAnswer":83},"What performance benefits does TWIN provide compared with existing approaches?",{"text":84,"@type":76},"TWIN transfers well across drug-like molecules, peptides, and proteins, matches DFT-based explicit-solvent MLP quality closely, and enables about a two-order-of-magnitude faster timestep evaluation.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]