Assessing Conformation Validity and Rationality of Deep Learning-Generated 3D Molecules

Assessing Conformation Validity and Rationality of Deep Learning-Generated 3D Molecules | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Assessing Conformation Validity and Rationality of Deep Learning-Generated 3D Molecules Bo Huang, Fan Fan, Bin Xi, Xianghu Meng, Han Wang, Bowen Zhang, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5479504/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 07 Feb, 2026 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Recent advancements in artificial intelligence (AI) have revolutionized the field of 3D molecule generation. However, the lack of effective evaluation methods for 3D conformations limits further improvements. Current techniques, in order to achieve the necessary speed for evaluating large number of AI-generated molecules, often rely on empirical geometric metrics that do not adequately capture various conformational anomalies, or on molecular mechanics (MM) energy metrics that exhibit low accuracy and lack atomic or torsional details. To address this gap, we propose a two-stage approach that achieves both high speed and quantum mechanical (QM) level accuracy. The first stage, termed the validity test, employs an AI-derived force field to identify atoms with elevated energy resulting from implausible neighboring environments. The second stage, known as the rationality test, utilizes a deep learning network trained on data with density functional theory (DFT) accuracy to detect rotatable bonds with high torsional energies. To demonstrate the functionality of our evaluation system, we applied our approach to five recently reported 3D molecule generation AI models across 102 targets in Directory of Useful Decoys-Enhanced (DUD-E) dataset. To facilitate accessibility for the academic community, our method is available as an open-source package. Biological sciences/Computational biology and bioinformatics/Machine learning Biological sciences/Drug discovery/Medicinal chemistry/Computational chemistry Biological sciences/Biological techniques/Software Full Text Additional Declarations There is NO Competing Interest. Supplementary Files ConformationPaperManuscriptSIv2.docx Supplementary Information machinelearningchecklistBH.pdf Machine Learning Checklist nrsoftwarepolicyBH.pdf Software Policy Checklist Cite Share Download PDF Status: Published Journal Publication published 07 Feb, 2026 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5479504","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":403137723,"identity":"6a6bc9ae-c258-4792-8992-d3074332fc16","order_by":0,"name":"Bo Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA30lEQVRIiWNgGAWjYBACAxDB2MAgB+VLEK/FmIGNVC2JDWzEOsxcIvnZw687bNLnz28+9uBjm4U9PwPzswf4tFjOSDM3lj2TlrvhGFu64cw2icSZDWzmBngddiPBTFqy7XDuBjYeM2nebRIJBgd42PD6yOBG+jeQlnT5NqCWv9sk7O0Ja8kxk/zYdjiB4RhQC+M2CcYNDIS0nHlTJs14Js1ww7G0NMnefxKJMw6zmeHXcjx9m+TPHTby8s2Hj0n8OFNnz9/e/Ixg7DDzoHIJqQcCxh9EKBoFo2AUjIIRDABAFUSp8ZwRKAAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-3822-9110","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":true,"prefix":"","firstName":"Bo","middleName":"","lastName":"Huang","suffix":""},{"id":403137724,"identity":"83e2af79-ccf8-40f6-af18-154ae97087f5","order_by":1,"name":"Fan Fan","email":"","orcid":"","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Fan","middleName":"","lastName":"Fan","suffix":""},{"id":403137725,"identity":"ccc86bcf-9873-4f6b-ab3a-7ac4e428d186","order_by":2,"name":"Bin Xi","email":"","orcid":"","institution":"State Key Laboratory of Natural and Biomimetic Drugs, School of Pharmaceutical Sciences, Peking University","correspondingAuthor":false,"prefix":"","firstName":"Bin","middleName":"","lastName":"Xi","suffix":""},{"id":403137726,"identity":"88b8e922-f646-4e8f-9fd3-6e31d038e7b6","order_by":3,"name":"Xianghu Meng","email":"","orcid":"","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Xianghu","middleName":"","lastName":"Meng","suffix":""},{"id":403137727,"identity":"54426523-9085-4caa-8b8f-34dbecb980ca","order_by":4,"name":"Han Wang","email":"","orcid":"https://orcid.org/0000-0001-9282-2105","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Han","middleName":"","lastName":"Wang","suffix":""},{"id":403137728,"identity":"7ce06d96-6a1c-4fbb-b7af-d78976ea3608","order_by":5,"name":"Bowen Zhang","email":"","orcid":"","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Bowen","middleName":"","lastName":"Zhang","suffix":""},{"id":403137729,"identity":"2ae52a5a-386e-4fa7-9205-2a2cdc7e60e4","order_by":6,"name":"Qingbo Xu","email":"","orcid":"","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Qingbo","middleName":"","lastName":"Xu","suffix":""},{"id":403137730,"identity":"b3a460df-a586-49dd-ba9b-d2c849da10dc","order_by":7,"name":"Wei Feng","email":"","orcid":"https://orcid.org/0009-0007-1220-1458","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Feng","suffix":""},{"id":403137731,"identity":"93f1ad0b-72db-43ff-be26-934fcc695608","order_by":8,"name":"Xiaoman Wang","email":"","orcid":"","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Xiaoman","middleName":"","lastName":"Wang","suffix":""},{"id":403137732,"identity":"9ca02c4c-a431-4115-a30d-c7112e408b21","order_by":9,"name":"Hongbo Zhang","email":"","orcid":"https://orcid.org/0009-0005-1780-1968","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Hongbo","middleName":"","lastName":"Zhang","suffix":""},{"id":403137733,"identity":"11f0f29f-c994-4e43-9909-cd738f47c162","order_by":10,"name":"Feng Zhou","email":"","orcid":"","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Feng","middleName":"","lastName":"Zhou","suffix":""},{"id":403137734,"identity":"2880679a-066f-44e9-832f-c9e5abfb7470","order_by":11,"name":"Zhenming Liu","email":"","orcid":"https://orcid.org/0000-0002-8993-4015","institution":"Peking University","correspondingAuthor":false,"prefix":"","firstName":"Zhenming","middleName":"","lastName":"Liu","suffix":""},{"id":403137735,"identity":"8d130028-ef8c-493f-b765-c6474b68af6d","order_by":12,"name":"Wenbiao Zhou","email":"","orcid":"","institution":"Beijing StoneWise Technology Co Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Wenbiao","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2024-11-19 02:45:18","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5479504/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5479504/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-026-69303-5","type":"published","date":"2026-02-07T05:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":104777018,"identity":"4e1a0776-a36b-484e-b2cd-a7cd5de93d22","added_by":"auto","created_at":"2026-03-17 07:05:43","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2260408,"visible":true,"origin":"","legend":"Article File","description":"","filename":"HEADTEDmanuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5479504/v1_covered_d60b3503-4f80-4e76-b06e-f1e23796be6f.pdf"},{"id":74201774,"identity":"066bea89-5d9b-41b3-b43f-8b422f795976","added_by":"auto","created_at":"2025-01-20 02:32:07","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":342851,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"ConformationPaperManuscriptSIv2.docx","url":"https://assets-eu.researchsquare.com/files/rs-5479504/v1/95d739f2779d218ec9a7f700.docx"},{"id":74201544,"identity":"c70cdc6b-01cd-41ab-a904-8ebefd6f5a71","added_by":"auto","created_at":"2025-01-20 02:24:07","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":216402,"visible":true,"origin":"","legend":"Machine Learning Checklist","description":"","filename":"machinelearningchecklistBH.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5479504/v1/16b6073a7a845ad17d023d77.pdf"},{"id":74201545,"identity":"9f13b6a1-6eaa-40e4-8604-a6bc5ac87dc4","added_by":"auto","created_at":"2025-01-20 02:24:08","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1317124,"visible":true,"origin":"","legend":"Software Policy Checklist","description":"","filename":"nrsoftwarepolicyBH.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5479504/v1/d9498523ab3544a8b21af1af.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Assessing Conformation Validity and Rationality of Deep Learning-Generated 3D Molecules","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5479504/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5479504/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eRecent advancements in artificial intelligence (AI) have revolutionized the field of 3D molecule generation. However, the lack of effective evaluation methods for 3D conformations limits further improvements. Current techniques, in order to achieve the necessary speed for evaluating large number of AI-generated molecules, often rely on empirical geometric metrics that do not adequately capture various conformational anomalies, or on molecular mechanics (MM) energy metrics that exhibit low accuracy and lack atomic or torsional details. To address this gap, we propose a two-stage approach that achieves both high speed and quantum mechanical (QM) level accuracy. The first stage, termed the validity test, employs an AI-derived force field to identify atoms with elevated energy resulting from implausible neighboring environments. The second stage, known as the rationality test, utilizes a deep learning network trained on data with density functional theory (DFT) accuracy to detect rotatable bonds with high torsional energies. To demonstrate the functionality of our evaluation system, we applied our approach to five recently reported 3D molecule generation AI models across 102 targets in Directory of Useful Decoys-Enhanced (DUD-E) dataset. To facilitate accessibility for the academic community, our method is available as an open-source package.\u003c/p\u003e","manuscriptTitle":"Assessing Conformation Validity and Rationality of Deep Learning-Generated 3D Molecules","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-01-20 02:24:03","doi":"10.21203/rs.3.rs-5479504/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"9868faab-00df-4275-98d6-ff9ee4f46eb9","owner":[],"postedDate":"January 20th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":42981363,"name":"Biological sciences/Computational biology and bioinformatics/Machine learning"},{"id":42981364,"name":"Biological sciences/Drug discovery/Medicinal chemistry/Computational chemistry"},{"id":42981365,"name":"Biological sciences/Biological techniques/Software"}],"tags":[],"updatedAt":"2026-03-17T07:05:35+00:00","versionOfRecord":{"articleIdentity":"rs-5479504","link":"https://doi.org/10.1038/s41467-026-69303-5","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2026-02-07 05:00:00","publishedOnDateReadable":"February 7th, 2026"},"versionCreatedAt":"2025-01-20 02:24:03","video":"","vorDoi":"10.1038/s41467-026-69303-5","vorDoiUrl":"https://doi.org/10.1038/s41467-026-69303-5","workflowStages":[]},"version":"v1","identity":"rs-5479504","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5479504","identity":"rs-5479504","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}