{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,19]],"date-time":"2026-03-19T06:34:34Z","timestamp":1773902074479,"version":"3.50.1"},"reference-count":32,"publisher":"Oxford University Press (OUP)","issue":"Supplement_2","license":[{"start":{"date-parts":[[2020,12,1]],"date-time":"2020-12-01T00:00:00Z","timestamp":1606780800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/academic.oup.com\/journals\/pages\/open_access\/funder_policies\/chorus\/standard_publication_model"}],"funder":[{"DOI":"10.13039\/501100002347","name":"BMBF","doi-asserted-by":"publisher","award":["01IS18036A"],"award-info":[{"award-number":["01IS18036A"]}],"id":[{"id":"10.13039\/501100002347","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100002347","name":"BMBF","doi-asserted-by":"publisher","award":["01IS18053A"],"award-info":[{"award-number":["01IS18053A"]}],"id":[{"id":"10.13039\/501100002347","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001659","name":"German Research Foundation","doi-asserted-by":"crossref","award":["ZT-I-0007"],"award-info":[{"award-number":["ZT-I-0007"]}],"id":[{"id":"10.13039\/501100001659","id-type":"DOI","asserted-by":"crossref"}]},{"name":"Chan Zuckerberg Initiative DAF","award":["182835"],"award-info":[{"award-number":["182835"]}]},{"DOI":"10.13039\/100008662","name":"Joachim Herz Stiftung","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100008662","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2020,12,30]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:sec>\n                  <jats:title>Motivation<\/jats:title>\n                  <jats:p>While generative models have shown great success in sampling high-dimensional samples conditional on low-dimensional descriptors (stroke thickness in MNIST, hair color in CelebA, speaker identity in WaveNet), their generation out-of-distribution poses fundamental problems due to the difficulty of learning compact joint distribution across conditions. The canonical example of the conditional variational autoencoder (CVAE), for instance, does not explicitly relate conditions during training and, hence, has no explicit incentive of learning such a compact representation.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>We overcome the limitation of the CVAE by matching distributions across conditions using maximum mean discrepancy in the decoder layer that follows the bottleneck. This introduces a strong regularization both for reconstructing samples within the same condition and for transforming samples across conditions, resulting in much improved generalization. As this amount to solving a style-transfer problem, we refer to the model as transfer VAE (trVAE). Benchmarking trVAE on high-dimensional image and single-cell RNA-seq, we demonstrate higher robustness and higher accuracy than existing approaches. We also show qualitatively improved predictions by tackling previously problematic minority classes and multiple conditions in the context of cellular perturbation response to treatment and disease based on high-dimensional single-cell gene expression data. For generic tasks, we improve Pearson correlations of high-dimensional estimated means and variances with their ground truths from 0.89 to 0.97 and 0.75 to 0.87, respectively. We further demonstrate that trVAE learns cell-type-specific responses after perturbation and improves the prediction of most cell-type-specific genes by 65%.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Availability and implementation<\/jats:title>\n                  <jats:p>The trVAE implementation is available via github.com\/theislab\/trvae. The results of this article can be reproduced via github.com\/theislab\/trvae_reproducibility.<\/jats:p>\n               <\/jats:sec>","DOI":"10.1093\/bioinformatics\/btaa800","type":"journal-article","created":{"date-parts":[[2020,10,19]],"date-time":"2020-10-19T19:15:09Z","timestamp":1603134909000},"page":"i610-i617","source":"Crossref","is-referenced-by-count":121,"title":["Conditional out-of-distribution generation for unpaired data using transfer VAE"],"prefix":"10.1093","volume":"36","author":[{"given":"Mohammad","family":"Lotfollahi","sequence":"first","affiliation":[{"name":"Institute of Computational Biology, Helmholtz Center Munich , Neuherberg, Germany"},{"name":"Technical University of Munich School of Life Sciences Weihenstephan, , Munich, Germany"}]},{"given":"Mohsen","family":"Naghipourfar","sequence":"additional","affiliation":[{"name":"Institute of Computational Biology, Helmholtz Center Munich , Neuherberg, Germany"}]},{"given":"Fabian J","family":"Theis","sequence":"additional","affiliation":[{"name":"Institute of Computational Biology, Helmholtz Center Munich , Neuherberg, Germany"},{"name":"Technical University of Munich School of Life Sciences Weihenstephan, , Munich, Germany"},{"name":"Technische Universit\u00e4t M\u00fcnchen Department of Mathematics, , Munich, Germany"}]},{"given":"F Alexander","family":"Wolf","sequence":"additional","affiliation":[{"name":"Institute of Computational Biology, Helmholtz Center Munich , Neuherberg, Germany"}]}],"member":"286","published-online":{"date-parts":[[2020,12,29]]},"reference":[{"key":"2023062409322048000_btaa800-B1","author":"Amodio","year":"2018"},{"key":"2023062409322048000_btaa800-B2","doi-asserted-by":"crossref","first-page":"1139","DOI":"10.1038\/s41592-019-0576-7","article-title":"Exploring single-cell data with deep multitasking neural networks","volume":"16","author":"Amodio","year":"2019","journal-title":"Nat. Methods"},{"key":"2023062409322048000_btaa800-B3","first-page":"214","volume-title":"Proceedings of the 34th International Conference on Machine Learning, Volume 70 of Proceedings of Machine Learning Research","author":"Arjovsky","year":"2017"},{"key":"2023062409322048000_btaa800-B4","author":"Bi\u0144kowski","year":"2018"},{"key":"2023062409322048000_btaa800-B5","author":"Castro","year":"2018"},{"key":"2023062409322048000_btaa800-B6","author":"Doersch","year":"2016"},{"key":"2023062409322048000_btaa800-B7","first-page":"258","volume-title":"Proceedings of the Thirty-First Conference on Uncertainty in Artificial Intelligence, UAI\u201915","author":"Dziugaite","year":"2015"},{"key":"2023062409322048000_btaa800-B8","author":"Dziugaite","year":"2015"},{"key":"2023062409322048000_btaa800-B9","doi-asserted-by":"crossref","first-page":"390","DOI":"10.1038\/s41467-018-07931-2","article-title":"Single-cell RNA-seq denoising using a deep count autoencoder","volume":"10","author":"Eraslan","year":"2019","journal-title":"Nat. Commun"},{"key":"2023062409322048000_btaa800-B10","first-page":"2672","volume-title":"Advances in Neural Information Processing Systems, Palais des Congr\u00e8s de Montr\u00e9al, Montr\u00e9al, Canada, pp.","author":"Goodfellow","year":"2014"},{"key":"2023062409322048000_btaa800-B11","first-page":"723","article-title":"A kernel two-sample test","volume":"13","author":"Gretton","year":"2012","journal-title":"J. Mach. Learn. Res"},{"key":"2023062409322048000_btaa800-B12","doi-asserted-by":"crossref","first-page":"333","DOI":"10.1038\/nature24489","article-title":"A single-cell survey of the small intestinal epithelium","volume":"551","author":"Haber","year":"2017","journal-title":"Nature"},{"key":"2023062409322048000_btaa800-B13","doi-asserted-by":"crossref","first-page":"3450","DOI":"10.1038\/s41598-020-59827-1","article-title":"Single-cell transcriptome mapping identifies common and cell-type specific genes affected by acute delta9-tetrahydrocannabinol in humans","volume":"10","author":"Hu","year":"2020","journal-title":"Sci. Rep"},{"key":"2023062409322048000_btaa800-B14","first-page":"3020","volume-title":"International Conference on Machine Learning","author":"Johansson","year":"2016"},{"key":"2023062409322048000_btaa800-B15","doi-asserted-by":"crossref","first-page":"89","DOI":"10.1038\/nbt.4042","article-title":"Multiplexed droplet single-cell RNA-sequencing using natural genetic variation","volume":"36","author":"Kang","year":"2018","journal-title":"Nat. Biotechnol"},{"key":"2023062409322048000_btaa800-B16","author":"Kingma","year":"2013"},{"key":"2023062409322048000_btaa800-B17","first-page":"1718","author":"Li","year":"2015"},{"key":"2023062409322048000_btaa800-B18","author":"Liu","year":"2015"},{"key":"2023062409322048000_btaa800-B19","author":"Long","year":"2015"},{"key":"2023062409322048000_btaa800-B20","doi-asserted-by":"crossref","first-page":"1053","DOI":"10.1038\/s41592-018-0229-2","article-title":"Deep generative modeling for single-cell transcriptomics","volume":"15","author":"Lopez","year":"2018","journal-title":"Nat. Methods"},{"key":"2023062409322048000_btaa800-B21","first-page":"6114","volume-title":"Advances in Neural Information Processing Systems","author":"Lopez","year":"2018"},{"key":"2023062409322048000_btaa800-B22","doi-asserted-by":"crossref","first-page":"715","DOI":"10.1038\/s41592-019-0494-8","article-title":"scGen predicts single-cell perturbation responses","volume":"16","author":"Lotfollahi","year":"2019","journal-title":"Nat. Methods"},{"key":"2023062409322048000_btaa800-B23","author":"Louizos","year":"2015"},{"key":"2023062409322048000_btaa800-B24","first-page":"6446","volume-title":"Advances in Neural Information Processing Systems","author":"Louizos","year":"2017"},{"key":"2023062409322048000_btaa800-B25","author":"McInnes","year":"2018"},{"key":"2023062409322048000_btaa800-B26","author":"Mirza","year":"2014"},{"key":"2023062409322048000_btaa800-B27","first-page":"2928","volume-title":"Advances in Neural Information Processing Systems","author":"Ren","year":"2016"},{"key":"2023062409322048000_btaa800-B28","first-page":"234","author":"Ronneberger","year":"2015"},{"key":"2023062409322048000_btaa800-B29","first-page":"3483","article-title":"Learning structured output representation using deep conditional generative models","volume":"Vol. 28","author":"Sohn","year":"2015","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2023062409322048000_btaa800-B30","doi-asserted-by":"crossref","first-page":"45","DOI":"10.1126\/science.aax6234","article-title":"Massively multiplex chemical transcriptomics at single-cell resolution","volume":"367","author":"Srivatsan","year":"2020","journal-title":"Science"},{"key":"2023062409322048000_btaa800-B31","author":"Tzeng","year":"2014"},{"key":"2023062409322048000_btaa800-B32","author":"Zhu","year":"2017"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/36\/Supplement_2\/i610\/50693361\/btaa800.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/36\/Supplement_2\/i610\/50693361\/btaa800.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,6,24]],"date-time":"2023-06-24T23:52:05Z","timestamp":1687650725000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/36\/Supplement_2\/i610\/6055927"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,12]]},"references-count":32,"journal-issue":{"issue":"Supplement_2","published-print":{"date-parts":[[2020,12,30]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btaa800","relation":{},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"value":"1367-4803","type":"print"},{"value":"1367-4811","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2020,12]]},"published":{"date-parts":[[2020,12]]}}}