Towards a universal foundation model for biobank-scale human genome variation

Towards a universal foundation model for biobank-scale human genome variation | 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 Towards a universal foundation model for biobank-scale human genome variation Augix Xu, Kun Tang, Yu Xu, Yiming Xing, Pengchao Luo, Jianbo Yang, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7855919/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Millions of human genomes have been genotyped by national biobanks worldwide. Training large language models (LLM) with this data may lead to a universal model of human genome with tremendous potential. Yet the quadrillions (10 15 ) of nucleotides—resulting from genome length multiplied by population size—pose formidable challenges for modeling. In this study, we propose a novel AI framework designed to scale with this data and support diverse analytical tasks. To demonstrate this scheme, we developed SNPBag—a foundation model focusing on single nucleotide polymorphism (SNP). With about 1 billion parameters, it is trained on one million synthesized human genomes, corresponding to a total of 6 trillion SNP tokens. SNPBag showed superior performance in benchmarking of multiple tasks. In genotype imputation, it achieves state-of-the-art (SOTA) accuracy. In haplotype phasing, it rivals the best method with a 72-fold speedup. By encoding 6 million SNPs per genome into a 0.75 MB embedding, SNPBag enables efficient storage, transfer and downstream applications. In particular, the genome embeddings facilitate rapid ancestry inference across global populations and detection of genetic relationships up to 12th-degree relatives. Collectively, SNPBag introduces a new paradigm for scalable, unified and multitask analysis of the ever-growing human variation data. Biological sciences/Genetics/Population genetics Biological sciences/Computational biology and bioinformatics/Software Physical sciences/Mathematics and computing/Computational science Full Text Additional Declarations There is NO Competing Interest. Cite Share Download PDF Status: Posted 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-7855919","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":542849246,"identity":"8e789cc2-2335-49e2-9c9c-aa438e6bd47b","order_by":0,"name":"Augix Xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAmUlEQVRIiWNgGAWjYJCCAwwVpGs5Q7I1jG2kqOafkX7xMO+8bXIM7IePbiBKi8SNnIKDM7fdNmbgSUu7QZw1N3ISDnzcdjuxQYLHjDgt8iAtiXNu1xOvxeBG+oEDHxtuJzAQrcXwzBuGgzOO3TZsI9ovcsfTH3/mqbktz89++BiR3mfgMQBTbEQqBwH2ByQoHgWjYBSMghEJAB6dNCYieR6AAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-8509-1137","institution":"Leipzig University","correspondingAuthor":true,"prefix":"","firstName":"Augix","middleName":"","lastName":"Xu","suffix":""},{"id":542849247,"identity":"7da8c2dc-eeee-4b84-96dd-bbd0aaf3953f","order_by":1,"name":"Kun Tang","email":"","orcid":"https://orcid.org/0009-0008-2330-1056","institution":"Zhejiang lab","correspondingAuthor":false,"prefix":"","firstName":"Kun","middleName":"","lastName":"Tang","suffix":""},{"id":542849248,"identity":"4269697b-ee94-4b59-9d00-646ffcc8c768","order_by":2,"name":"Yu Xu","email":"","orcid":"","institution":"BGI","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Xu","suffix":""},{"id":542849249,"identity":"69407d1d-f263-4251-ac25-09ecd8d96463","order_by":3,"name":"Yiming Xing","email":"","orcid":"","institution":"Zhejiang University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Yiming","middleName":"","lastName":"Xing","suffix":""},{"id":542849250,"identity":"c3e4bf8a-d084-4c12-af45-a979bd5fca7e","order_by":4,"name":"Pengchao Luo","email":"","orcid":"","institution":"University of Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Pengchao","middleName":"","lastName":"Luo","suffix":""},{"id":542849251,"identity":"03f08edf-2553-47e3-aa7b-98ee46028a15","order_by":5,"name":"Jianbo Yang","email":"","orcid":"https://orcid.org/0000-0001-7965-1110","institution":"Chinese Academy of Sciences","correspondingAuthor":false,"prefix":"","firstName":"Jianbo","middleName":"","lastName":"Yang","suffix":""},{"id":542849252,"identity":"0eb8dbd8-d300-41c8-889e-79c02df2cc1b","order_by":6,"name":"Yinqi Bai","email":"","orcid":"https://orcid.org/0000-0003-1017-5712","institution":"BGI Researh","correspondingAuthor":false,"prefix":"","firstName":"Yinqi","middleName":"","lastName":"Bai","suffix":""}],"badges":[],"createdAt":"2025-10-14 08:20:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7855919/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7855919/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104923431,"identity":"0aa66b23-f9d8-4703-bb9f-41eff8d59cbc","added_by":"auto","created_at":"2026-03-18 18:11:10","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2786057,"visible":true,"origin":"","legend":"Article File","description":"","filename":"SNPBagNG.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7855919/v1_covered_e657da62-d8b0-43fe-9b18-0bcd05232f24.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Towards a universal foundation model for biobank-scale human genome variation","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7855919/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7855919/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Millions of human genomes have been genotyped by national biobanks worldwide. Training large language models (LLM) with this data may lead to a universal model of human genome with tremendous potential. Yet the quadrillions (10\u003csup\u003e15\u003c/sup\u003e) of nucleotides—resulting from genome length multiplied by population size—pose formidable challenges for modeling. In this study, we propose a novel AI framework designed to scale with this data and support diverse analytical tasks. To demonstrate this scheme, we developed SNPBag—a foundation model focusing on single nucleotide polymorphism (SNP). With about 1 billion parameters, it is trained on one million synthesized human genomes, corresponding to a total of 6 trillion SNP tokens. SNPBag showed superior performance in benchmarking of multiple tasks. In genotype imputation, it achieves state-of-the-art (SOTA) accuracy. In haplotype phasing, it rivals the best method with a 72-fold speedup. By encoding 6 million SNPs per genome into a 0.75 MB embedding, SNPBag enables efficient storage, transfer and downstream applications. In particular, the genome embeddings facilitate rapid ancestry inference across global populations and detection of genetic relationships up to 12th-degree relatives. Collectively, SNPBag introduces a new paradigm for scalable, unified and multitask analysis of the ever-growing human variation data.","manuscriptTitle":"Towards a universal foundation model for biobank-scale human genome variation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-11 11:14:29","doi":"10.21203/rs.3.rs-7855919/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5596f040-fcf6-4f63-a115-572ed93c4587","owner":[],"postedDate":"November 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":57748772,"name":"Biological sciences/Genetics/Population genetics"},{"id":57748773,"name":"Biological sciences/Computational biology and bioinformatics/Software"},{"id":57748774,"name":"Physical sciences/Mathematics and computing/Computational science"}],"tags":[],"updatedAt":"2026-03-18T18:10:52+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-11 11:14:29","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7855919","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7855919","identity":"rs-7855919","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}