Evaluating the learnability of single-cell large language models on multiple tasks

Evaluating the learnability of single-cell large language models on multiple tasks | 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 Research Article Evaluating the learnability of single-cell large language models on multiple tasks Yu Yan, Xutao Wang, Dongyuan Song This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8919408/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 13 You are reading this latest preprint version Abstract The rise of single-cell foundation models (scFMs) has sparked interest in their potential to unify diverse biological tasks. However, their practical utility and the validity of scaling laws—the assumption that performance improves with model and data size—remain under-examined. Here, we systematically evaluate two representative scFMs, Geneformer and scGPT, across perturbation prediction and cell type annotation tasks. Our findings suggest that the benefits of large-scale pretraining are strongly task-dependent, conferring substantial advantages in cell type annotation but limited gains in perturbation prediction. Furthermore, our results indicate that increasing model size does not guarantee improved performance and can even be detrimental, challenging the ``bigger is better'' paradigm. By comparing model performance on real versus synthetic data with different levels of complexity, our analysis suggests that for perturbation prediction, the tested scFMs capture little more than simple summary statistics and may struggle to learn complex biological interactions. These results highlight the need to move beyond scaling and toward developing models that integrate deeper biological knowledge. We suggest that a renewed focus on task-specific architectures and biologically-informed priors may be critical for unlocking the true potential of foundation models in single-cell biology. Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 23 Mar, 2026 Reviews received at journal 22 Mar, 2026 Reviews received at journal 10 Mar, 2026 Reviewers agreed at journal 09 Mar, 2026 Reviewers agreed at journal 08 Mar, 2026 Reviewers agreed at journal 08 Mar, 2026 Reviews received at journal 07 Mar, 2026 Reviewers agreed at journal 07 Mar, 2026 Reviewers invited by journal 06 Mar, 2026 Editor invited by journal 05 Mar, 2026 Editor assigned by journal 02 Mar, 2026 Submission checks completed at journal 02 Mar, 2026 First submitted to journal 19 Feb, 2026 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-8919408","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":603750953,"identity":"6a583bda-953c-462d-82b5-6df4c6b3ee93","order_by":0,"name":"Yu Yan","email":"","orcid":"","institution":"University of Connecticut Health Center","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Yan","suffix":""},{"id":603750954,"identity":"2bb0c88d-a5aa-4c5f-9caa-a8552f452c67","order_by":1,"name":"Xutao Wang","email":"","orcid":"","institution":"Rutgers University–New Jersey Medical School","correspondingAuthor":false,"prefix":"","firstName":"Xutao","middleName":"","lastName":"Wang","suffix":""},{"id":603750955,"identity":"58feabce-e454-4ef4-b9ba-63da80b1a4c8","order_by":2,"name":"Dongyuan Song","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAq0lEQVRIiWNgGAWjYDCCA2DShoFBAkSzEa8ljXQth0nQwne8/eHjgl/nE/tnNz9g+FB2mLAWyTNnjI1n9t1OnHHnmAHjjHNEaDG4kcMmzdtzO3GDRA4DM28bUVrSnwG1nINo+UuclgQzaZ4fByBaGInRAvYLb0Oy8YwbaQYHe86lE9YCDjGeP3ay/TOSHz74UWZNWAsYMLZB6ANEqgeBPySoHQWjYBSMgpEHAL8RPwW+/q1xAAAAAElFTkSuQmCC","orcid":"","institution":"University of Connecticut Health Center","correspondingAuthor":true,"prefix":"","firstName":"Dongyuan","middleName":"","lastName":"Song","suffix":""}],"badges":[],"createdAt":"2026-02-19 16:38:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8919408/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8919408/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104780411,"identity":"71266b97-ae23-4ef9-b024-a86f3cadbf4d","added_by":"auto","created_at":"2026-03-17 07:52:50","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":8431363,"visible":true,"origin":"","legend":"","description":"","filename":"scLLMBMCGenomics02272026.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8919408/v1_covered_8d423398-7b15-4bcb-bee7-941c11405371.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Evaluating the learnability of single-cell large language models on multiple tasks","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"bmc-genomics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"gics","sideBox":"Learn more about [BMC Genomics](http://bmcgenomics.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/gics","title":"BMC Genomics","twitterHandle":"#BMCGenomics","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8919408/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8919408/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe rise of single-cell foundation models (scFMs) has sparked interest in their potential to unify diverse biological tasks. However, their practical utility and the validity of scaling laws\u0026mdash;the assumption that performance improves with model and data size\u0026mdash;remain under-examined. Here, we systematically evaluate two representative scFMs, Geneformer and scGPT, across perturbation prediction and cell type annotation tasks. Our findings suggest that the benefits of large-scale pretraining are strongly task-dependent, conferring substantial advantages in cell type annotation but limited gains in perturbation prediction. Furthermore, our results indicate that increasing model size does not guarantee improved performance and can even be detrimental, challenging the ``bigger is better'' paradigm. By comparing model performance on real versus synthetic data with different levels of complexity, our analysis suggests that for perturbation prediction, the tested scFMs capture little more than simple summary statistics and may struggle to learn complex biological interactions. These results highlight the need to move beyond scaling and toward developing models that integrate deeper biological knowledge. We suggest that a renewed focus on task-specific architectures and biologically-informed priors may be critical for unlocking the true potential of foundation models in single-cell biology.\u003c/p\u003e","manuscriptTitle":"Evaluating the learnability of single-cell large language models on multiple tasks","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-12 06:01:01","doi":"10.21203/rs.3.rs-8919408/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-03-23T10:24:38+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-22T14:53:55+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-10T11:15:48+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"78391462975996674835193563812715410823","date":"2026-03-09T09:28:07+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"144068601863278125746497792298514060433","date":"2026-03-09T02:42:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"138150541896897871197814319654154130047","date":"2026-03-08T16:38:58+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-07T22:00:34+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"195146070248285907517056360184278330325","date":"2026-03-07T20:28:42+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-03-06T15:13:53+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-03-05T06:11:56+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-02T22:36:55+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-02T22:35:02+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Genomics","date":"2026-02-19T16:22:06+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-genomics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"gics","sideBox":"Learn more about [BMC Genomics](http://bmcgenomics.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/gics","title":"BMC Genomics","twitterHandle":"#BMCGenomics","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"2cce6a72-f0b1-47bc-874c-e89e7092aa72","owner":[],"postedDate":"March 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-15T22:23:41+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-12 06:01:01","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8919408","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8919408","identity":"rs-8919408","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}