{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,4]],"date-time":"2026-08-04T15:16:34Z","timestamp":1785856594316,"version":"3.56.0"},"reference-count":18,"publisher":"Oxford University Press (OUP)","issue":"17","license":[{"start":{"date-parts":[[2018,9,9]],"date-time":"2018-09-09T00:00:00Z","timestamp":1536451200000},"content-version":"vor","delay-in-days":8,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"BioX Stanford Interdisciplinary Graduate Fellowship"},{"name":"SIGF"},{"name":"National Science Foundation Graduate Research Fellowship"},{"name":"National Institute of Health","award":["1DP2GM123485"],"award-info":[{"award-number":["1DP2GM123485"]}]},{"name":"National Institute of Health","award":["1U01HG009431"],"award-info":[{"award-number":["1U01HG009431"]}]},{"name":"National Institute of Health","award":["1R01HG00967401"],"award-info":[{"award-number":["1R01HG00967401"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2018,9,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Motivation<\/jats:title>\n                    <jats:p>Transcription factors bind regulatory DNA sequences in a combinatorial manner to modulate gene expression. Deep neural networks (DNNs) can learn the cis-regulatory grammars encoded in regulatory DNA sequences associated with transcription factor binding and chromatin accessibility. Several feature attribution methods have been developed for estimating the predictive importance of individual features (nucleotides or motifs) in any input DNA sequence to its associated output prediction from a DNN model. However, these methods do not reveal higher-order feature interactions encoded by the models.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>We present a new method called Deep Feature Interaction Maps (DFIM) to efficiently estimate interactions between all pairs of features in any input DNA sequence. DFIM accurately identifies ground truth motif interactions embedded in simulated regulatory DNA sequences. DFIM identifies synergistic interactions between GATA1 and TAL1 motifs from in\u00a0vivo TF binding models. DFIM reveals epistatic interactions involving nucleotides flanking the core motif of the Cbf1 TF in yeast from in\u00a0vitro TF binding models. We also apply DFIM to regulatory sequence models of in\u00a0vivo chromatin accessibility to reveal interactions between regulatory genetic variants and proximal motifs of target TFs as validated by TF binding quantitative trait loci. Our approach makes significant strides in improving the interpretability of deep learning models for genomics.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Availability and implementation<\/jats:title>\n                    <jats:p>Code is available at: https:\/\/github.com\/kundajelab\/dfim.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Supplementary information<\/jats:title>\n                    <jats:p>Supplementary data are available at Bioinformatics online.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1093\/bioinformatics\/bty575","type":"journal-article","created":{"date-parts":[[2018,7,7]],"date-time":"2018-07-07T02:00:37Z","timestamp":1530928837000},"page":"i629-i637","source":"Crossref","is-referenced-by-count":85,"title":["Discovering epistatic feature interactions from neural network models of regulatory DNA sequences"],"prefix":"10.1093","volume":"34","author":[{"given":"Peyton","family":"Greenside","sequence":"first","affiliation":[{"name":"Biomedical Informatics Training Program, Stanford University, Stanford, CA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tyler","family":"Shimko","sequence":"additional","affiliation":[{"name":"Genetics, Stanford University, Stanford, CA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Polly","family":"Fordyce","sequence":"additional","affiliation":[{"name":"Genetics, Stanford University, Stanford, CA"},{"name":"Bioengineering, Stanford University, Stanford, CA"},{"name":"Chan Zuckerberg Biohub, San Francisco, CA, USA"},{"name":"Chem-H Institute, Stanford University, Stanford, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Anshul","family":"Kundaje","sequence":"additional","affiliation":[{"name":"Genetics, Stanford University, Stanford, CA"},{"name":"Computer Science, Stanford University, Stanford, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2018,9,8]]},"reference":[{"key":"2023061313495609600_bty575-B1","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":"2023061313495609600_bty575-B2","doi-asserted-by":"crossref","first-page":"782","DOI":"10.1038\/s41467-018-03082-6","article-title":"Exploiting genetic variation to uncover rules of transcription factor binding and chromatin accessibility","volume":"9","author":"Behera","year":"2018","journal-title":"Nat. Commun."},{"key":"2023061313495609600_bty575-B3","doi-asserted-by":"crossref","first-page":"1193","DOI":"10.1038\/ng.3646","article-title":"Lineage-specific and single-cell chromatin accessibility charts human hematopoiesis and leukemia evolution","volume":"48","author":"Corces","year":"2016","journal-title":"Nat. Genet."},{"key":"2023061313495609600_bty575-B4","doi-asserted-by":"crossref","first-page":"57","DOI":"10.1038\/nature11247","article-title":"An integrated encyclopedia of dna elements in the human genome","volume":"489","author":"ENCODE Project Consortium","year":"2012","journal-title":"Nature"},{"key":"2023061313495609600_bty575-B5","doi-asserted-by":"crossref","first-page":"e1005836","DOI":"10.1371\/journal.pcbi.1005836","article-title":"Maximum entropy methods for extracting the learned features of deep neural networks","volume":"13","author":"Finnegan","year":"2017","journal-title":"PLoS Comput. Biol."},{"key":"2023061313495609600_bty575-B6","doi-asserted-by":"crossref","first-page":"1189","DOI":"10.1214\/aos\/1013203451","article-title":"Greedy function approximation: a gradient boosting machine","volume":"29","author":"Friedman","year":"2001","journal-title":"Ann. Stat."},{"key":"2023061313495609600_bty575-B7","doi-asserted-by":"crossref","first-page":"91","DOI":"10.1038\/nature11245","article-title":"Architecture of the human regulatory network derived from ENCODE data","volume":"489","author":"Gerstein","year":"2012","journal-title":"Nature"},{"key":"2023061313495609600_bty575-B8","doi-asserted-by":"crossref","first-page":"1064","DOI":"10.1101\/gr.104935.110","article-title":"Genome-wide identification of TAL1\u2019s functional targets: insights into its mechanisms of action in primary erythroid cells","volume":"20","author":"Kassouf","year":"2010","journal-title":"Genome Res."},{"key":"2023061313495609600_bty575-B9","doi-asserted-by":"crossref","first-page":"990","DOI":"10.1101\/gr.200535.115","article-title":"Basset: learning the regulatory code of the accessible genome with deep convolutional neural networks","volume":"26","author":"Kelley","year":"2016","journal-title":"Genome Res"},{"key":"2023061313495609600_bty575-B10","doi-asserted-by":"crossref","first-page":"2976","DOI":"10.1093\/nar\/gkt1249","article-title":"Systematic discovery and characterization of regulatory motifs in ENCODE TF binding experiments","volume":"42","author":"Kheradpour","year":"2014","journal-title":"Nucleic Acids Res."},{"key":"2023061313495609600_bty575-B11","doi-asserted-by":"crossref","first-page":"E3702","DOI":"10.1073\/pnas.1715888115","article-title":"Comprehensive, high-resolution binding energy landscapes reveal context dependencies of transcription factor binding","volume":"115","author":"Le","year":"2018","journal-title":"Proc. Natl. Acad. Sci. USA"},{"key":"2023061313495609600_bty575-B12","first-page":"4765","article-title":"A unified approach to interpreting model predictions","volume-title":"Advances in Neural Information Processing Systems","author":"Lundberg","year":"2017"},{"key":"2023061313495609600_bty575-B13","first-page":"9","article-title":"Consistent individualized feature attribution for tree ensembles","author":"Lundberg","year":"2018","journal-title":"Proceedings of ACM (KDD' 18)"},{"key":"2023061313495609600_bty575-B14","first-page":"3145","article-title":"Learning important features through propagating activation differences","volume-title":"Proceedings of the 34th International Conference on Machine Learning","author":"Shrikumar","year":"2017"},{"key":"2023061313495609600_bty575-B15","article-title":"Deep inside convolutional networks: visualising image classification models and saliency maps","author":"Simonyan","year":"2014","journal-title":"Workshop Proceedings of the International Conference on Learning Representations (ICLR)"},{"key":"2023061313495609600_bty575-B16","first-page":"3319","article-title":"Axiomatic attribution for deep networks","author":"Sundararajan","year":"2017","journal-title":"Proceedings of the 34th International Conference on Machine Learning, PMLR 70"},{"key":"2023061313495609600_bty575-B17","doi-asserted-by":"crossref","first-page":"730","DOI":"10.1016\/j.cell.2016.03.041","article-title":"Pooled ChIP-seq links variation in transcription factor binding to complex disease risk","volume":"165","author":"Tehranchi","year":"2016","journal-title":"Cell"},{"key":"2023061313495609600_bty575-B18","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"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/34\/17\/i629\/50582291\/bioinformatics_34_17_i629.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/34\/17\/i629\/50582291\/bioinformatics_34_17_i629.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,6,13]],"date-time":"2023-06-13T09:51:15Z","timestamp":1686649875000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/34\/17\/i629\/5093210"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2018,9,1]]},"references-count":18,"journal-issue":{"issue":"17","published-print":{"date-parts":[[2018,9,1]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/bty575","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/302711","asserted-by":"object"}]},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"value":"1367-4803","type":"print"},{"value":"1367-4811","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2018,9,1]]},"published":{"date-parts":[[2018,9,1]]}}}