{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,26]],"date-time":"2026-02-26T20:33:35Z","timestamp":1772138015837,"version":"3.50.1"},"reference-count":42,"publisher":"Oxford University Press (OUP)","issue":"1","license":[{"start":{"date-parts":[[2021,9,28]],"date-time":"2021-09-28T00:00:00Z","timestamp":1632787200000},"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":"Bundesministerium f\u00fcr Bildung und Forschung","doi-asserted-by":"publisher","award":["031L0205A"],"award-info":[{"award-number":["031L0205A"]}],"id":[{"id":"10.13039\/501100002347","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2022,1,17]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Deep neural networks are frequently employed to predict survival conditional on omics-type biomarkers, e.g., by employing the partial likelihood of Cox proportional hazards model as loss function. Due to the generally limited number of observations in clinical studies, combining different data sets has been proposed to improve learning of network parameters. However, if baseline hazards differ between the studies, the assumptions of Cox proportional hazards model are violated. Based on high dimensional transcriptome profiles from different tumor entities, we demonstrate how using a stratified partial likelihood as loss function allows for accounting for the different baseline hazards in a deep learning framework. Additionally, we compare the partial likelihood with the ranking loss, which is frequently employed as loss function in machine learning approaches due to its seemingly simplicity. Using RNA-seq data from the Cancer Genome Atlas (TCGA) we show that use of stratified loss functions leads to an overall better discriminatory power and lower prediction error compared to their non-stratified counterparts. We investigate which genes are identified to have the greatest marginal impact on prediction of survival when using different loss functions. We find that while similar genes are identified, in particular known prognostic genes receive higher importance from stratified loss functions. Taken together, pooling data from different sources for improved parameter learning of deep neural networks benefits largely from employing stratified loss functions that consider potentially varying baseline hazards. For easy application, we provide PyTorch code for stratified loss functions and an explanatory Jupyter notebook in a GitHub repository.<\/jats:p>","DOI":"10.1093\/bib\/bbab392","type":"journal-article","created":{"date-parts":[[2021,9,8]],"date-time":"2021-09-08T07:20:51Z","timestamp":1631085651000},"source":"Crossref","is-referenced-by-count":3,"title":["Stratified neural networks in a time-to-event setting"],"prefix":"10.1093","volume":"23","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-3685-1005","authenticated-orcid":false,"given":"Fabrizio","family":"Kuruc","sequence":"first","affiliation":[{"name":"Institute of Medical Biometry and Statistics, Faculty of Medicine and Medical Center - University of Freiburg, Germany"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5666-8662","authenticated-orcid":false,"given":"Harald","family":"Binder","sequence":"additional","affiliation":[{"name":"Institute of Medical Biometry and Statistics, Faculty of Medicine and Medical Center - University of Freiburg, Germany"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4021-1796","authenticated-orcid":false,"given":"Moritz","family":"Hess","sequence":"additional","affiliation":[{"name":"Institute of Medical Biometry and Statistics, Faculty of Medicine and Medical Center - University of Freiburg, Germany"}]}],"member":"286","published-online":{"date-parts":[[2021,9,28]]},"reference":[{"key":"2022011921121684700_ref1","doi-asserted-by":"crossref","first-page":"217","DOI":"10.1056\/NEJMoa063994","article-title":"The prognostic role of a gene signature from tumorigenic breast-cancer cells","volume":"356","author":"Liu","year":"2007","journal-title":"N Engl J Med"},{"key":"2022011921121684700_ref2","doi-asserted-by":"crossref","first-page":"717","DOI":"10.1056\/NEJMoa1602253","article-title":"70-gene signature as an aid to treatment decisions in early-stage breast cancer","volume":"375","author":"Cardoso","year":"2016","journal-title":"N Engl J Med"},{"key":"2022011921121684700_ref3","doi-asserted-by":"crossref","first-page":"841","DOI":"10.1214\/08-AOAS169","article-title":"Random survival forests","volume":"2","author":"Ishwaran","year":"2008","journal-title":"Ann Appl Stat"},{"key":"2022011921121684700_ref4","first-page":"1","volume-title":"Proceedings of the third international conference on computational intelligence in medicine and healthcare (cimed2007)","author":"Van Belle","year":"2007"},{"key":"2022011921121684700_ref5","doi-asserted-by":"crossref","first-page":"14","DOI":"10.1186\/1471-2105-9-14","article-title":"Allowing for mandatory covariates in boosting estimation of sparse high-dimensional survival models","volume":"9","author":"Binder","year":"2008","journal-title":"BMC Bioinformatics"},{"key":"2022011921121684700_ref6","doi-asserted-by":"crossref","first-page":"e84483","DOI":"10.1371\/journal.pone.0084483","article-title":"Boosting the concordance index for survival data\u2013a unified framework to derive and evaluate biomarker combinations","volume":"9","author":"Mayr","year":"2014","journal-title":"PLoS One"},{"key":"2022011921121684700_ref7","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1038\/s41598-017-11817-6","article-title":"Predicting clinical outcomes from large scale cancer genomic profiles with deep survival models","volume":"7","author":"Yousefi","year":"2017","journal-title":"Sci Rep"},{"key":"2022011921121684700_ref8","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1016\/j.artmed.2019.06.001","article-title":"A deep survival analysis method based on ranking","volume":"98","author":"Jing","year":"2019","journal-title":"Artif Intell Med"},{"key":"2022011921121684700_ref9","doi-asserted-by":"crossref","first-page":"1248","DOI":"10.1158\/1078-0432.CCR-17-0853","article-title":"Deep learning\u2013based multi-omics integration robustly predicts survival in liver cancer","volume":"24","author":"Chaudhary","year":"2018","journal-title":"Clin Cancer Res"},{"key":"2022011921121684700_ref10","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1371\/journal.pcbi.1006076","article-title":"Cox-nnet: an artificial neural network method for prognosis prediction of high-throughput omics data","volume":"14","author":"Ching","year":"2018","journal-title":"PLoS Comput Biol"},{"key":"2022011921121684700_ref11","doi-asserted-by":"crossref","first-page":"187","DOI":"10.1111\/j.2517-6161.1972.tb00899.x","article-title":"Regression models and life-tables","volume":"34","author":"Cox","year":"1972","journal-title":"J R Stat Soc B Methodol"},{"key":"2022011921121684700_ref12","doi-asserted-by":"crossref","first-page":"24","DOI":"10.1186\/s12874-018-0482-1","article-title":"DeepSurv: personalized treatment recommender system using a Cox proportional hazards deep neural network","volume":"18","author":"Katzman","year":"2018","journal-title":"BMC Med Res Methodol"},{"key":"2022011921121684700_ref13","doi-asserted-by":"crossref","first-page":"289","DOI":"10.1093\/biomet\/74.2.289","article-title":"A simple test of the proportional hazards assumption","volume":"74","author":"Gill","year":"1987","journal-title":"Biometrika"},{"key":"2022011921121684700_ref14","doi-asserted-by":"crossref","first-page":"1707","DOI":"10.1002\/sim.4780141510","article-title":"Graphical methods for assessing violations of the proportional hazards assumption in cox regression","volume":"14","author":"Hess","year":"1995","journal-title":"Stat Med"},{"key":"2022011921121684700_ref15","first-page":"227","article-title":"Proportional hazard regression models and the analysis of censored survival data","volume":"26","author":"Kay","year":"1977","journal-title":"J R Stat Soc Ser C Appl Stat"},{"key":"2022011921121684700_ref16","doi-asserted-by":"crossref","first-page":"2543","DOI":"10.1001\/jama.1982.03320430047030","article-title":"Evaluating the yield of medical tests","volume":"247","author":"Harrell","year":"1982","journal-title":"JAMA"},{"key":"2022011921121684700_ref17","first-page":"1209","volume-title":"Proceedings of the 20th International Conference on Neural Information Processing Systems","author":"Raykar","year":"2007"},{"key":"2022011921121684700_ref18","doi-asserted-by":"crossref","first-page":"85","DOI":"10.1016\/0895-4356(92)90192-P","article-title":"What price perfection? Calibration and discrimination of clinical prediction models","volume":"45","author":"Diamond","year":"1992","journal-title":"J Clin Epidemiol"},{"key":"2022011921121684700_ref19","doi-asserted-by":"crossref","first-page":"2529","DOI":"10.1002\/(SICI)1097-0258(19990915\/30)18:17\/18<2529::AID-SIM274>3.0.CO;2-5","article-title":"Assessment and comparison of prognostic classification schemes for survival data","volume":"18","author":"Graf","year":"1999","journal-title":"Stat Med"},{"key":"2022011921121684700_ref20","doi-asserted-by":"crossref","first-page":"359","DOI":"10.1016\/0893-6080(89)90020-8","article-title":"Multilayer feedforward networks are universal approximators","volume":"2","author":"Hornik","year":"1989","journal-title":"Neural Netw"},{"key":"2022011921121684700_ref21","doi-asserted-by":"crossref","first-page":"533","DOI":"10.1038\/323533a0","article-title":"Learning representations by back-propagating errors","volume":"323","author":"Rumelhart","year":"1986","journal-title":"Nature"},{"key":"2022011921121684700_ref22","doi-asserted-by":"crossref","first-page":"275","DOI":"10.1002\/bimj.201000145","article-title":"Measures of prediction error for survival data with longitudinal covariates","volume":"53","author":"Schoop","year":"2011","journal-title":"Biom J"},{"key":"2022011921121684700_ref23","doi-asserted-by":"crossref","first-page":"610","DOI":"10.1002\/bimj.200800157","article-title":"Classification of therapy resistance based on longitudinal biomarker profiles","volume":"51","author":"Kohlmann","year":"2009","journal-title":"Biometrical Journal Biometrische Zeitschrift"},{"key":"2022011921121684700_ref24","doi-asserted-by":"crossref","first-page":"1029","DOI":"10.1002\/bimj.200610301","article-title":"Consistent estimation of the expected Brier score in general survival models with right-censored event times","volume":"48","author":"Gerds","year":"2006","journal-title":"Biom J"},{"issue":"477","key":"2022011921121684700_ref25","doi-asserted-by":"crossref","first-page":"359","DOI":"10.1198\/016214506000001437","article-title":"Strictly proper scoring rules, prediction, and estimation","volume":"102","author":"Gneiting","year":"2007","journal-title":"J Am Stat Assoc"},{"key":"2022011921121684700_ref26","doi-asserted-by":"crossref","first-page":"1113","DOI":"10.1038\/ng.2764","article-title":"The cancer genome atlas pan-cancer analysis project","volume":"45","author":"Weinstein","year":"2013","journal-title":"Nat Genet"},{"key":"2022011921121684700_ref27","doi-asserted-by":"crossref","first-page":"3666","DOI":"10.1093\/bioinformatics\/btv377","article-title":"Alternative preprocessing of RNA-sequencing data in the cancer genome atlas leads to improved analysis results","volume":"31","author":"Rahman","year":"2015","journal-title":"Bioinformatics"},{"key":"2022011921121684700_ref28","doi-asserted-by":"crossref","first-page":"9546","DOI":"10.1073\/pnas.0914005107","article-title":"Independent filtering increases detection power for high-throughput experiments","volume":"107","author":"Bourgon","year":"2010","journal-title":"Proc Natl Acad Sci"},{"key":"2022011921121684700_ref29","doi-asserted-by":"crossref","first-page":"9","DOI":"10.1007\/978-3-642-35289-8_3","volume-title":"Neural Networks: Tricks of the Trade","author":"LeCun","year":"2012"},{"key":"2022011921121684700_ref30","article-title":"Adam: a method for stochastic optimization","volume-title":"arXiv","author":"Kingma","year":"2014"},{"issue":"5","key":"2022011921121684700_ref31","doi-asserted-by":"crossref","first-page":"1","DOI":"10.18637\/jss.v039.i05","article-title":"Regularization paths for Cox\u2019s proportional hazards model via coordinate descent","volume":"39","author":"Simon","year":"2011","journal-title":"J Stat Softw"},{"key":"2022011921121684700_ref32","first-page":"4765","article-title":"A unified approach to interpreting model predictions","volume":"30","author":"Lundberg","year":"2017","journal-title":"Adv Neural Inf Proces Syst"},{"issue":"28","key":"2022011921121684700_ref33","first-page":"307","article-title":"A value for n-person games","volume":"2","author":"Shapley","year":"1953","journal-title":"Contributions to the Theory of Games"},{"key":"2022011921121684700_ref34","first-page":"3145","article-title":"Learning important features through propagating activation differences.","volume-title":"Proc Int Conf Mach Learn","author":"Shrikumar","year":"2017"},{"key":"2022011921121684700_ref35","doi-asserted-by":"crossref","first-page":"24","DOI":"10.1186\/s40659-018-0172-9","article-title":"Knockdown of paics inhibits malignant proliferation of human breast cancer cell lines","volume":"51","author":"Meng","year":"2018","journal-title":"Biol Res"},{"key":"2022011921121684700_ref36","doi-asserted-by":"crossref","first-page":"223","DOI":"10.1016\/j.yexcr.2015.02.005","article-title":"Two members of the tric chaperonin complex, cct2 and tcp1 are essential for survival of breast cancer cells and are linked to driving oncogenes","volume":"332","author":"Guest","year":"2015","journal-title":"Exp Cell Res"},{"key":"2022011921121684700_ref37","doi-asserted-by":"crossref","first-page":"112029","DOI":"10.1016\/j.yexcr.2020.112029","article-title":"Knockdown of alpk2 blocks development and progression of renal cell carcinoma","volume":"392","author":"Jiang","year":"2020","journal-title":"Exp Cell Res"},{"key":"2022011921121684700_ref38","doi-asserted-by":"crossref","first-page":"1261","DOI":"10.1158\/1078-0432.CCR-18-2312","article-title":"Cell surface notch ligand dll3 is a therapeutic target in isocitrate dehydrogenase\u2013mutant glioma","volume":"25","author":"Spino","year":"2019","journal-title":"Clin Cancer Res"},{"key":"2022011921121684700_ref39","doi-asserted-by":"crossref","first-page":"319","DOI":"10.1593\/neo.05682","article-title":"ADAM15 disintegrin is associated with aggressive prostate and breast cancer disease","volume":"8","author":"Kuefer","year":"2006","journal-title":"Neoplasia (New York, NY)"},{"issue":"Supplement_1","key":"2022011921121684700_ref40","doi-asserted-by":"crossref","first-page":"i389","DOI":"10.1093\/bioinformatics\/btaa462","article-title":"Improved survival analysis by learning shared genomic information from pan-cancer data","volume":"36","author":"Kim","year":"2020","journal-title":"Bioinformatics"},{"issue":"8","key":"2022011921121684700_ref41","first-page":"1","article-title":"Combining gene ontology with deep neural networks to enhance the clustering of single cell RNA-Seq data","volume":"20","author":"Peng","year":"2019","journal-title":"BMC Bioinformatics"},{"key":"2022011921121684700_ref42","doi-asserted-by":"crossref","first-page":"533","DOI":"10.1080\/00949655.2017.1397151","article-title":"Statistical power to detect violation of the proportional hazards assumption when using the cox regression model","volume":"88","author":"Austin","year":"2018","journal-title":"J Stat Comput Simul"}],"container-title":["Briefings in Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bib\/article-pdf\/23\/1\/bbab392\/42230671\/bbab392.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bib\/article-pdf\/23\/1\/bbab392\/42230671\/bbab392.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,9,7]],"date-time":"2024-09-07T21:04:32Z","timestamp":1725743072000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bib\/article\/doi\/10.1093\/bib\/bbab392\/6377517"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,9,28]]},"references-count":42,"journal-issue":{"issue":"1","published-print":{"date-parts":[[2022,1,17]]}},"URL":"https:\/\/doi.org\/10.1093\/bib\/bbab392","relation":{"has-preprint":[{"id-type":"doi","id":"10.1101\/2021.02.01.429169","asserted-by":"object"}]},"ISSN":["1467-5463","1477-4054"],"issn-type":[{"value":"1467-5463","type":"print"},{"value":"1477-4054","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2022,1]]},"published":{"date-parts":[[2021,9,28]]},"article-number":"bbab392"}}