{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,24]],"date-time":"2026-04-24T15:15:25Z","timestamp":1777043725022,"version":"3.51.4"},"reference-count":27,"publisher":"Oxford University Press (OUP)","issue":"4","license":[{"start":{"date-parts":[[2026,3,24]],"date-time":"2026-03-24T00:00:00Z","timestamp":1774310400000},"content-version":"vor","delay-in-days":1,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"NIH\/NIMH","award":["R00MH136290"],"award-info":[{"award-number":["R00MH136290"]}]},{"DOI":"10.13039\/100000001","name":"NSF","doi-asserted-by":"publisher","award":["2211718"],"award-info":[{"award-number":["2211718"]}],"id":[{"id":"10.13039\/100000001","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,4,7]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Motivation<\/jats:title>\n                    <jats:p>Recent studies suggest that deep neural network models trained on thousands of human genomic datasets can accurately predict genomic features, including gene expression and chromatin accessibility. However, training these models is computation- and time-intensive, and datasets of comparable size do not exist for most other organisms.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>Here, we identify modifications to an existing state-of-the-art model that improve model accuracy while reducing training time and computational cost. Using this streamlined model architecture, we investigate the ability of models pretrained on human genomic datasets to transfer performance to a variety of different tasks. Models pretrained on human data but fine-tuned on genomic datasets from diverse tissues and species achieved significantly higher prediction accuracy while significantly reducing training time compared to models trained from scratch, with Pearson correlation coefficients between experimental results and predictions as high as 0.8. Further, we found that including excessive training tasks decreased model performance and that this decrease could be partially but not completely rescued by fine-tuning. Thus, simplifying model architecture, applying pretrained models, and carefully considering the number of training tasks may be effective and economical techniques for building new models across data types, tissues, and species.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Availability and implementation<\/jats:title>\n                    <jats:p>Code is available on GitHub and Figshare: https:\/\/github.com\/optimizedlearning\/genomicsML, https:\/\/doi.org\/10.6084\/m9.figshare.31796116.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1093\/bioinformatics\/btag139","type":"journal-article","created":{"date-parts":[[2026,3,19]],"date-time":"2026-03-19T12:43:10Z","timestamp":1773924190000},"source":"Crossref","is-referenced-by-count":0,"title":["Pretraining improves prediction of genomic datasets across species"],"prefix":"10.1093","volume":"42","author":[{"given":"Fangrui","family":"Huang","sequence":"first","affiliation":[{"name":"Boston University Department of Computer Science, , Boston, MA, 02215,","place":["USA"]},{"name":"Stanford University Department of Computer Science, , Stanford, CA, 94305,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yitong","family":"Wang","sequence":"additional","affiliation":[{"name":"Boston University Department of Computer Science, , Boston, MA, 02215,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ashok","family":"Cutkosky","sequence":"additional","affiliation":[{"name":"Boston University Department of Electrical and Computer Engineering, , Boston, MA, 02215,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7898-0227","authenticated-orcid":false,"given":"Janet H T","family":"Song","sequence":"additional","affiliation":[{"name":"Harvard University Department of Human Evolutionary Biology, , Cambridge, MA, 02138,","place":["USA"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"286","published-online":{"date-parts":[[2026,3,23]]},"reference":[{"key":"2026042409462463800_btag139-B1","author":"Abbas","year":"2023"},{"key":"2026042409462463800_btag139-B2","doi-asserted-by":"publisher","first-page":"107663","DOI":"10.1016\/j.celrep.2020.107663","article-title":"Predicting mRNA abundance directly from genomic sequence using deep convolutional neural networks","volume":"31","author":"Agarwal","year":"2020","journal-title":"Cell Rep"},{"key":"2026042409462463800_btag139-B3","doi-asserted-by":"crossref","first-page":"831","DOI":"10.1038\/nbt.3300","article-title":"Predicting the sequence specificities of DNA-and RNA-binding proteins by deep learning","volume":"33","author":"Alipanahi","year":"2015","journal-title":"Nat Biotechnol"},{"key":"2026042409462463800_btag139-B4","doi-asserted-by":"crossref","first-page":"1196","DOI":"10.1038\/s41592-021-01252-x","article-title":"Effective gene expression prediction from sequence by integrating long-range interactions","volume":"18","author":"Avsec","year":"2021","journal-title":"Nat Methods"},{"key":"2026042409462463800_btag139-B5","doi-asserted-by":"crossref","first-page":"354","DOI":"10.1038\/s41588-021-00782-6","article-title":"Base-resolution models of transcription-factor binding reveal soft motif syntax","volume":"53","author":"Avsec","year":"2021","journal-title":"Nat Genet"},{"key":"2026042409462463800_btag139-B6","doi-asserted-by":"publisher","author":"Devlin","year":"2019","DOI":"10.18653\/v1\/N19-1423"},{"key":"2026042409462463800_btag139-B7","author":"Dosovitskiy","year":"2021"},{"key":"2026042409462463800_btag139-B8","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1186\/s12864-020-07078-9","article-title":"A comparative analysis of chromatin accessibility in cattle, pig, and mouse tissues","volume":"21","author":"Halstead","year":"2020","journal-title":"BMC Genomics"},{"key":"2026042409462463800_btag139-B9","author":"Hernandez","year":"2021"},{"key":"2026042409462463800_btag139-B10","author":"Hu","year":"2021"},{"key":"2026042409462463800_btag139-B11","doi-asserted-by":"crossref","first-page":"e1008050","DOI":"10.1371\/journal.pcbi.1008050","article-title":"Cross-species regulatory sequence activity prediction","volume":"16","author":"Kelley","year":"2020","journal-title":"PLoS Comput Biol"},{"key":"2026042409462463800_btag139-B12","doi-asserted-by":"crossref","first-page":"739","DOI":"10.1101\/gr.227819.117","article-title":"Sequential regulatory activity prediction across chromosomes with convolutional neural networks","volume":"28","author":"Kelley","year":"2018","journal-title":"Genome Res"},{"key":"2026042409462463800_btag139-B13","author":"Kern","year":"2021"},{"key":"2026042409462463800_btag139-B14","author":"Linder","year":"2023"},{"key":"2026042409462463800_btag139-B15","doi-asserted-by":"crossref","first-page":"949","DOI":"10.1038\/s41588-024-02053-6","article-title":"Predicting RNA-seq coverage from DNA sequence as a unifying model of gene regulation","volume":"57","author":"Linder","year":"2025","journal-title":"Nat Genet"},{"key":"2026042409462463800_btag139-B16","author":"Loshchilov","year":"2017"},{"key":"2026042409462463800_btag139-B17","first-page":"43177","author":"Nguyen"},{"key":"2026042409462463800_btag139-B18","first-page":"8026","article-title":"Pytorch: an imperative style, high-performance deep learning library","volume":"32","author":"Paszke","year":"2019","journal-title":"Adv Neural Inform Process Syst"},{"key":"2026042409462463800_btag139-B19","first-page":"9","article-title":"Language models are unsupervised multitask learners","volume":"1","author":"Radford","year":"2019","journal-title":"OpenAI Blog"},{"key":"2026042409462463800_btag139-B20","doi-asserted-by":"crossref","first-page":"272","DOI":"10.1038\/nrg3682","article-title":"Transcriptional enhancers: from properties to genome-wide predictions","volume":"15","author":"Shlyueva","year":"2014","journal-title":"Nat Rev Genet"},{"key":"2026042409462463800_btag139-B21","first-page":"6000","article-title":"Attention is all you need","volume":"30","author":"Vaswani","year":"2017","journal-title":"Adv Neural Inform Process Syst"},{"key":"2026042409462463800_btag139-B22","doi-asserted-by":"crossref","first-page":"494","DOI":"10.1038\/nn.4229","article-title":"Epigenomic annotation of gene regulatory alterations during evolution of the primate brain","volume":"19","author":"Vermunt","year":"2016","journal-title":"Nat Neurosci"},{"key":"2026042409462463800_btag139-B23","doi-asserted-by":"crossref","DOI":"10.1017\/9781108627771","volume-title":"High-Dimensional Statistics: A Non-Asymptotic Viewpoint","author":"Wainwright","year":"2019"},{"key":"2026042409462463800_btag139-B24","first-page":"3320","article-title":"How transferable are features in deep neural networks?","volume":"27","author":"Yosinski","year":"2014","journal-title":"Adv Neural Inf Process Syst"},{"key":"2026042409462463800_btag139-B25","doi-asserted-by":"publisher","author":"Yuan","year":"2025","DOI":"10.1101\/2025.05.26.656171"},{"key":"2026042409462463800_btag139-B26","doi-asserted-by":"crossref","first-page":"931","DOI":"10.1038\/nmeth.3547","article-title":"Predicting effects of noncoding variants with deep learning-based sequence model","volume":"12","author":"Zhou","year":"2015","journal-title":"Nat Methods"},{"key":"2026042409462463800_btag139-B27","doi-asserted-by":"crossref","first-page":"e247","DOI":"10.1371\/journal.pcbi.0030247","article-title":"Comparative genomics search for losses of long-established genes on the human lineage","volume":"3","author":"Zhu","year":"2007","journal-title":"PLoS Comput Biol"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/advance-article-pdf\/doi\/10.1093\/bioinformatics\/btag139\/67486017\/btag139.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/42\/4\/btag139\/67486017\/btag139.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/42\/4\/btag139\/67486017\/btag139.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,24]],"date-time":"2026-04-24T13:46:37Z","timestamp":1777038397000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/doi\/10.1093\/bioinformatics\/btag139\/8538004"}},"subtitle":[],"editor":[{"given":"Peter","family":"Robinson","sequence":"additional","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2026,3,23]]},"references-count":27,"journal-issue":{"issue":"4","published-print":{"date-parts":[[2026,4,7]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btag139","relation":{},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"value":"1367-4803","type":"print"},{"value":"1367-4811","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,4]]},"published":{"date-parts":[[2026,3,23]]},"article-number":"btag139"}}