Multi-Modal Large Language Model Enables Protein Function Prediction

Multi-Modal Large Language Model Enables Protein Function Prediction | 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 Multi-Modal Large Language Model Enables Protein Function Prediction Pengtao Xie, Mingjia Huo, Han Guo, Xingyi Cheng, Digvijay Singh, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4941886/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 Predicting the functions of proteins can greatly accelerate biological discovery and applications, where deep learning methods have recently shown great potential. However, these methods predominantly predict protein functions as discrete categories, which fails to capture the nuanced and complex nature of protein functions. Furthermore, existing methods require the development of separate models for each prediction task, a process that can be both resource-heavy and time-consuming. Here, we present ProteinChat, a versatile, multi-modal large language model that takes a protein's amino acid sequence as input and generates comprehensive narratives describing its function. ProteinChat is trained using over 1,500,000 (protein, prompt, answer) triplets curated from the Swiss-Prot dataset, covering diverse functions. This novel model can universally predict a wide range of protein functions, all within a single, unified framework. Furthermore, ProteinChat supports interactive dialogues with human users, allowing for iterative refinement of predictions and deeper exploration of protein functions. Our experimental results, evaluated through both human expert assessment and automated metrics, demonstrate that ProteinChat outperforms general-purpose LLMs like GPT-4, one of the flagship LLMs, by over ten-fold. In addition, ProteinChat exceeds or matches the performance of task-specific prediction models. Biological sciences/Computational biology and bioinformatics/Protein function predictions Biological sciences/Computational biology and bioinformatics/Machine learning Protein Function Prediction Large Language Models Multi-modal Learning 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-4941886","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":351604319,"identity":"9c8086a5-ebe4-491c-a6ba-de1b47e42567","order_by":0,"name":"Pengtao Xie","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsklEQVRIiWNgGAWjYDACZgaGAwwVIFYCAwMPUVrYQVrOkKSFH4gZ20jRYs7MY3jg47zD8vLtCYwP3rYRocWymS3h4Mxthw03nHnAbDiXGC0Gh5kPHObddjjBQCKBTZqXOC2MDYf/zjmcID8jgf03kVqAtgB1JTDcSGBjJkoL2C89x9KBfnnYLDnnHBFazPnPGH/4UWMNDLHkgx/elBHjMASTsYEI9ahaRsEoGAWjYBTgAAD/TDf/lAJyMwAAAABJRU5ErkJggg==","orcid":"","institution":"University of California San Diego","correspondingAuthor":true,"prefix":"","firstName":"Pengtao","middleName":"","lastName":"Xie","suffix":""},{"id":351604320,"identity":"c8b7d8d1-ba3d-4d79-bb87-9c619a6e1d00","order_by":1,"name":"Mingjia Huo","email":"","orcid":"https://orcid.org/0000-0003-1636-3933","institution":"University of California San Diego","correspondingAuthor":false,"prefix":"","firstName":"Mingjia","middleName":"","lastName":"Huo","suffix":""},{"id":351604321,"identity":"622cb62f-113f-4807-912b-f92a46c72ff4","order_by":2,"name":"Han Guo","email":"","orcid":"","institution":"University of California San Diego","correspondingAuthor":false,"prefix":"","firstName":"Han","middleName":"","lastName":"Guo","suffix":""},{"id":351604322,"identity":"417e41c8-6b52-4b4a-904f-173c34658269","order_by":3,"name":"Xingyi Cheng","email":"","orcid":"","institution":"BioMap Research","correspondingAuthor":false,"prefix":"","firstName":"Xingyi","middleName":"","lastName":"Cheng","suffix":""},{"id":351604323,"identity":"a1ef3c55-f8e8-4149-b7d7-418d15ac9e18","order_by":4,"name":"Digvijay Singh","email":"","orcid":"","institution":"University of California San Diego","correspondingAuthor":false,"prefix":"","firstName":"Digvijay","middleName":"","lastName":"Singh","suffix":""},{"id":351604324,"identity":"63d6ff20-6d12-47ac-9f9d-18f0c385d1a9","order_by":5,"name":"Hamidreza Rahmani","email":"","orcid":"","institution":"The Scripps Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Hamidreza","middleName":"","lastName":"Rahmani","suffix":""},{"id":351604325,"identity":"49659fd4-95a6-476f-a940-ffd062899979","order_by":6,"name":"Shen Li","email":"","orcid":"","institution":"BioMap Research","correspondingAuthor":false,"prefix":"","firstName":"Shen","middleName":"","lastName":"Li","suffix":""},{"id":351604326,"identity":"fe6b0b6c-fe3f-420a-9116-c7a3f9fd12c4","order_by":7,"name":"Philipp Gerlof","email":"","orcid":"","institution":"The Scripps Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Philipp","middleName":"","lastName":"Gerlof","suffix":""},{"id":351604327,"identity":"3227cd72-15a7-410e-83fe-f1e56854fc19","order_by":8,"name":"Trey Ideker","email":"","orcid":"","institution":"University of California, San Diego","correspondingAuthor":false,"prefix":"","firstName":"Trey","middleName":"","lastName":"Ideker","suffix":""},{"id":351604328,"identity":"c5d731de-1779-4633-b62b-61f6f901569e","order_by":9,"name":"Danielle Grotjahn","email":"","orcid":"https://orcid.org/0000-0001-5908-7882","institution":"The Scripps Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Danielle","middleName":"","lastName":"Grotjahn","suffix":""},{"id":351604329,"identity":"2ea894ba-0cce-4965-a940-f60c56e46a08","order_by":10,"name":"Elizabeth Villa","email":"","orcid":"https://orcid.org/0000-0003-4677-9809","institution":"University of California San Diego","correspondingAuthor":false,"prefix":"","firstName":"Elizabeth","middleName":"","lastName":"Villa","suffix":""},{"id":351604330,"identity":"68b60dca-4728-4ac0-84a8-020bac91ca64","order_by":11,"name":"Le Song","email":"","orcid":"https://orcid.org/0000-0002-9655-2787","institution":"BioMap","correspondingAuthor":false,"prefix":"","firstName":"Le","middleName":"","lastName":"Song","suffix":""}],"badges":[],"createdAt":"2024-08-20 04:05:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4941886/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4941886/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":74000416,"identity":"62070f11-423d-47be-96db-4b62edc848de","added_by":"auto","created_at":"2025-01-16 21:19:11","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2129391,"visible":true,"origin":"","legend":"Article File","description":"","filename":"manuscriptproteinchat.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4941886/v1_covered_fc3ad7f5-a4a9-4ae8-9e62-a2e3149b8b89.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Multi-Modal Large Language Model Enables Protein Function Prediction","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":"Protein Function Prediction, Large Language Models, Multi-modal Learning","lastPublishedDoi":"10.21203/rs.3.rs-4941886/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4941886/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Predicting the functions of proteins can greatly accelerate biological discovery and applications, where deep learning methods have recently shown great potential. However, these methods predominantly predict protein functions as discrete categories, which fails to capture the nuanced and complex nature of protein functions. Furthermore, existing methods require the development of separate models for each prediction task, a process that can be both resource-heavy and time-consuming. Here, we present ProteinChat, a versatile, multi-modal large language model that takes a protein's amino acid sequence as input and generates comprehensive narratives describing its function. ProteinChat is trained using over 1,500,000 (protein, prompt, answer) triplets curated from the Swiss-Prot dataset, covering diverse functions. This novel model can universally predict a wide range of protein functions, all within a single, unified framework. Furthermore, ProteinChat supports interactive dialogues with human users, allowing for iterative refinement of predictions and deeper exploration of protein functions. Our experimental results, evaluated through both human expert assessment and automated metrics, demonstrate that ProteinChat outperforms general-purpose LLMs like GPT-4, one of the flagship LLMs, by over ten-fold. In addition, ProteinChat exceeds or matches the performance of task-specific prediction models.","manuscriptTitle":"Multi-Modal Large Language Model Enables Protein Function Prediction","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-09-10 03:25:20","doi":"10.21203/rs.3.rs-4941886/v1","editorialEvents":[],"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":"eac2dcdb-bbc1-409f-8e31-6e23b3fc8d56","owner":[],"postedDate":"September 10th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":37333618,"name":"Biological sciences/Computational biology and bioinformatics/Protein function predictions"},{"id":37333619,"name":"Biological sciences/Computational biology and bioinformatics/Machine learning"}],"tags":[],"updatedAt":"2025-05-12T17:45:29+00:00","versionOfRecord":[],"versionCreatedAt":"2024-09-10 03:25:20","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4941886","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4941886","identity":"rs-4941886","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}