{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,28]],"date-time":"2026-08-28T18:16:22Z","timestamp":1787940982557,"version":"build-2784847793"},"reference-count":45,"publisher":"Oxford University Press (OUP)","issue":"8","license":[{"start":{"date-parts":[[2026,7,26]],"date-time":"2026-07-26T00:00:00Z","timestamp":1785024000000},"content-version":"vor","delay-in-days":1,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"name":"SciLifeLab & Wallenberg Data-Driven Life Science Program","award":["KAW 2020.0239"],"award-info":[{"award-number":["KAW 2020.0239"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,8,3]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Motivation<\/jats:title>\n                    <jats:p>Computational models of cellular networks hold promise to uncover disease mechanisms and guide therapeutic strategies. Biology-informed neural networks (BINNs) is an emerging approach to create such models by combining the predictive power of deep learning with prior knowledge, a vital aspect of biological research. The architectures of BINN\u2019s enforces a network structure from which mechanism can ideally be inferred. However, a key challenge is to evaluate the reliability of these models, as cells are inherently complex, involving intricate and sometimes unknown interactions. Currently, analysis mainly focuses on selected pathways rather than a more comprehensive perspective.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>In this work we demonstrate an alternative holistic approach: we measure to which extent purposefully introduced spurious interactions are down-weighted by a BINN during training (self-pruning). The metric suggested RRA (Relative Residual Area) allows for direct distribution comparison with perfect self-pruning achieved at zero and a failure to self-prune if above one. To enable rapid testing, we updated LEMBAS (Large-scale knowledge-EMBedded Artificial Signaling-networks), our recurrent neural network framework for intracellular signaling dynamics, with full GPU acceleration. Our implementation achieves a &amp;gt;\u20097-fold speedup compared to the original while preserving predictive accuracy. We evaluated self-pruning in 3 different datasets and found that when spurious interactions are introduced at random, the model prunes these to a larger extent than those from the prior knowledge network (PKN), provided the model is regularized with a sufficiently large L2 norm. This suggests that BINNs can be robust to uncertainty in the PKN.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Availability and Implementation<\/jats:title>\n                    <jats:p>Our implementation of LEMBAS is freely available under a MIT license at https:\/\/github.com\/AvlantNilssonLab\/LEMBAS_GPU. The models and results to generate the figures can be downloaded through https:\/\/zenodo.org\/records\/17425598.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1093\/bioinformatics\/btag556","type":"journal-article","created":{"date-parts":[[2026,7,24]],"date-time":"2026-07-24T11:40:10Z","timestamp":1784893210000},"source":"Crossref","is-referenced-by-count":0,"title":["Biologically informed neural network models are robust to spurious interactions via self-pruning"],"prefix":"10.1093","volume":"42","author":[{"given":"Olof","family":"Nordenstorm","sequence":"first","affiliation":[{"name":"Department of Cell and Molecular Biology, Karolinska Institutet , Stockholm, 17177,","place":["Sweden"]},{"name":"Science for Life Laboratory (SciLifeLab), Karolinska Institutet , Stockholm, 17177,","place":["Sweden"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hratch","family":"Baghdassarian","sequence":"additional","affiliation":[{"name":"Department of Biological Engineering, Massachusetts Institute of Technology , Cambridge, MA 02139,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1850-0946","authenticated-orcid":false,"given":"Xuechun","family":"Xu","sequence":"additional","affiliation":[{"name":"Department of Cell and Molecular Biology, Karolinska Institutet , Stockholm, 17177,","place":["Sweden"]},{"name":"Science for Life Laboratory (SciLifeLab), Karolinska Institutet , Stockholm, 17177,","place":["Sweden"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0050-989X","authenticated-orcid":false,"given":"Douglas A","family":"Lauffenburger","sequence":"additional","affiliation":[{"name":"Department of Biological Engineering, Massachusetts Institute of Technology , Cambridge, MA 02139,","place":["United States"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9476-4516","authenticated-orcid":false,"given":"Avlant","family":"Nilsson","sequence":"additional","affiliation":[{"name":"Department of Cell and Molecular Biology, Karolinska Institutet , Stockholm, 17177,","place":["Sweden"]},{"name":"Science for Life Laboratory (SciLifeLab), Karolinska Institutet , Stockholm, 17177,","place":["Sweden"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2026,7,25]]},"reference":[{"key":"2026082813185013400_btag556-B1","doi-asserted-by":"crossref","first-page":"20170237","DOI":"10.1098\/rsif.2017.0237","article-title":"How to deal with parameters for whole-cell modelling","volume":"14","author":"Babtie","year":"2017","journal-title":"J R Soc Interface"},{"key":"2026082813185013400_btag556-B2","doi-asserted-by":"crossref","first-page":"7045","DOI":"10.1016\/j.cell.2024.11.015","article-title":"How to build the virtual cell with artificial intelligence: priorities and opportunities","volume":"187","author":"Bunne","year":"2024","journal-title":"Cell"},{"key":"2026082813185013400_btag556-B3","doi-asserted-by":"crossref","first-page":"88","DOI":"10.1007\/s10915-022-01939-z","article-title":"Scientific machine learning through physics\u2013informed neural networks: where we are and what\u2019s next","volume":"92","author":"Cuomo","year":"2022","journal-title":"J Sci Comput"},{"key":"2026082813185013400_btag556-B4","doi-asserted-by":"crossref","first-page":"348","DOI":"10.1038\/s41586-021-03922-4","article-title":"Biologically informed deep neural network for prostate cancer discovery","volume":"598","author":"Elmarakeby","year":"2021","journal-title":"Nature"},{"key":"2026082813185013400_btag556-B5","author":"Frankle"},{"key":"2026082813185013400_btag556-B6","doi-asserted-by":"crossref","first-page":"5359","DOI":"10.1038\/s41467-023-41146-4","article-title":"Interpreting biologically informed neural networks for enhanced proteomic biomarker discovery and pathway analysis","volume":"14","author":"Hartman","year":"2023","journal-title":"Nat Commun"},{"key":"2026082813185013400_btag556-B7","doi-asserted-by":"crossref","first-page":"194","DOI":"10.1002\/wsbm.52","article-title":"Computational modeling of mammalian signaling networks","volume":"2","author":"Hughey","year":"2010","journal-title":"Wiley Interdiscip Rev Syst Biol Med"},{"key":"2026082813185013400_btag556-B8","doi-asserted-by":"crossref","first-page":"297","DOI":"10.1038\/nrg2750","article-title":"Towards genome-scale signalling network reconstructions","volume":"11","author":"Hyduke","year":"2010","journal-title":"Nat Rev Genet"},{"key":"2026082813185013400_btag556-B9","doi-asserted-by":"crossref","first-page":"525","DOI":"10.1042\/EBC20180019","article-title":"Reconstructing phosphorylation signalling networks from quantitative phosphoproteomic data","volume":"62","author":"Invergo","year":"2018","journal-title":"Essays Biochem"},{"key":"2026082813185013400_btag556-B10","doi-asserted-by":"crossref","first-page":"27","DOI":"10.1093\/nar\/28.1.27","article-title":"KEGG: kyoto encyclopedia of genes and genomes","volume":"28","author":"Kanehisa","year":"2000","journal-title":"Nucleic Acids Res"},{"key":"2026082813185013400_btag556-B11","first-page":"58","article-title":"Challenges in the calibration of large-scale ordinary differential equation models","volume":"52","author":"Kapfer","year":"2019","journal-title":"IFAC-Pap"},{"key":"2026082813185013400_btag556-B12","doi-asserted-by":"crossref","first-page":"165","DOI":"10.1038\/nrm1838","article-title":"Cell-signalling dynamics in time and space","volume":"7","author":"Kholodenko","year":"2006","journal-title":"Nat Rev Mol Cell Biol"},{"key":"2026082813185013400_btag556-B13","doi-asserted-by":"crossref","first-page":"1126156","DOI":"10.3389\/fnagi.2023.1126156","article-title":"PINNet: a deep neural network with pathway prior knowledge for Alzheimer\u2019s disease","volume":"15","author":"Kim","year":"2023","journal-title":"Front Aging Neurosci"},{"key":"2026082813185013400_btag556-B14","author":"Kipf"},{"key":"2026082813185013400_btag556-B15","doi-asserted-by":"crossref","first-page":"281","DOI":"10.1186\/s13059-019-1887-9","article-title":"Guidelines for benchmarking of optimization-based approaches for fitting mathematical models","volume":"20","author":"Kreutz","year":"2019","journal-title":"Genome Biol"},{"key":"2026082813185013400_btag556-B16","article-title":"Towards understanding grokking: an effective theory of representation learning","volume":"35","author":"","year":"2022","journal-title":"Adv Neural Inf Process Syst"},{"key":"2026082813185013400_btag556-B17","year":"2023"},{"key":"2026082813185013400_btag556-B18","doi-asserted-by":"crossref","first-page":"btae316","DOI":"10.1093\/bioinformatics\/btae316","article-title":"Pathformer: a biological pathway informed transformer for disease diagnosis and prognosis using multi-omics data","volume":"40","author":"Liu","year":"2024","journal-title":"Bioinformatics"},{"key":"2026082813185013400_btag556-B19","doi-asserted-by":"crossref","first-page":"262","DOI":"10.1186\/1471-2105-8-262","article-title":"False positive reduction in protein-protein interaction predictions using gene ontology annotations","volume":"8","author":"Mahdavi","year":"2007","journal-title":"BMC Bioinformatics"},{"key":"2026082813185013400_btag556-B20","doi-asserted-by":"crossref","first-page":"424","DOI":"10.1016\/j.chaos.2005.09.075","article-title":"A new stability criterion for discrete-time neural networks: nonlinear spectral radius","volume":"31","author":"Mak","year":"2007","journal-title":"Chaos Solitons Fractals"},{"key":"2026082813185013400_btag556-B21","doi-asserted-by":"crossref","first-page":"109509","DOI":"10.1016\/j.isci.2024.109509","article-title":"Inference of drug off-target effects on cellular signaling using interactome-based deep learning","volume":"27","author":"Meimetis","year":"2024","journal-title":"iScience"},{"key":"2026082813185013400_btag556-B22","year":"2021"},{"key":"2026082813185013400_btag556-B23"},{"key":"2026082813185013400_btag556-B24"},{"key":"2026082813185013400_btag556-B25","author":"Neelakantan","year":"2015"},{"key":"2026082813185013400_btag556-B26","doi-asserted-by":"crossref","first-page":"40","DOI":"10.1145\/1365490.1365500","article-title":"Scalable parallel programming with CUDA","volume":"6","author":"Nickolls","year":"2008","journal-title":"Queue"},{"key":"2026082813185013400_btag556-B27","doi-asserted-by":"crossref","first-page":"3069","DOI":"10.1038\/s41467-022-30684-y","article-title":"Artificial neural networks enable genome-scale simulations of intracellular signaling","volume":"13","author":"Nilsson","year":"2022","journal-title":"Nat Commun"},{"key":"2026082813185013400_btag556-B28","doi-asserted-by":"crossref","first-page":"46","DOI":"10.1038\/s41698-025-00822-y","article-title":"Towards an interpretable deep learning model of cancer","volume":"9","author":"Nilsson","year":"2025","journal-title":"NPJ Precis Oncol"},{"key":"2026082813185013400_btag556-B29","author":"Power","year":"2022"},{"key":"2026082813185013400_btag556-B30","doi-asserted-by":"crossref","first-page":"093039","DOI":"10.1088\/1367-2630\/18\/9\/093039","article-title":"Optimal network modification for spectral radius dependent phase transitions","volume":"18","author":"Rosen","year":"2016","journal-title":"New J Phys"},{"key":"2026082813185013400_btag556-B31","doi-asserted-by":"crossref","first-page":"43","DOI":"10.1186\/1478-811X-11-43","article-title":"Modeling approaches for qualitative and semi-quantitative analysis of cellular signaling networks","volume":"11","author":"Samaga","year":"2013","journal-title":"Cell Commun Signal"},{"key":"2026082813185013400_btag556-B32","doi-asserted-by":"crossref","first-page":"1728","DOI":"10.1038\/s41467-022-29268-7","article-title":"Current progress and open challenges for applying deep learning across the biosciences","volume":"13","author":"Sapoval","year":"2022","journal-title":"Nat Commun"},{"key":"2026082813185013400_btag556-B33","doi-asserted-by":"crossref","first-page":"371","DOI":"10.1038\/s41576-025-00826-1","article-title":"Beyond the black box with biologically informed neural networks","volume":"26","author":"Selby","year":"2025","journal-title":"Nat Rev Genet"},{"key":"2026082813185013400_btag556-B34","author":"Sengupta","year":"2018"},{"key":"2026082813185013400_btag556-B35","doi-asserted-by":"crossref","first-page":"a006098","DOI":"10.1101\/cshperspect.a006098","article-title":"Signal transduction in cancer","volume":"5","author":"Sever","year":"2015","journal-title":"Cold Spring Harb Perspect Med"},{"key":"2026082813185013400_btag556-B36","doi-asserted-by":"crossref","first-page":"214","DOI":"10.1038\/s44320-025-00088-3","article-title":"Leveraging prior knowledge to infer gene regulatory networks from single-cell RNA-sequencing data","volume":"21","author":"Stock","year":"2025","journal-title":"Mol Syst Biol"},{"key":"2026082813185013400_btag556-B37","first-page":"3319","year":"2017"},{"key":"2026082813185013400_btag556-B38","doi-asserted-by":"crossref","first-page":"910","DOI":"10.1016\/j.cell.2016.03.029","article-title":"Oncogenic KRAS regulates tumor cell signaling via stromal reciprocation","volume":"165","author":"Tape","year":"2016","journal-title":"Cell"},{"key":"2026082813185013400_btag556-B39","doi-asserted-by":"crossref","first-page":"966","DOI":"10.1038\/nmeth.4077","article-title":"OmniPath: guidelines and gateway for literature-curated signaling pathway resources","volume":"13","author":"T\u00fcrei","year":"2016","journal-title":"Nat Methods"},{"key":"2026082813185013400_btag556-B40","author":"Veli\u010dkovi\u0107"},{"key":"2026082813185013400_btag556-B41","doi-asserted-by":"crossref","first-page":"261","DOI":"10.1038\/s41592-019-0686-2","article-title":"SciPy 1.0: fundamental algorithms for scientific computing in python","volume":"17","author":"Virtanen","year":"2020","journal-title":"Nat Methods"},{"key":"2026082813185013400_btag556-B42","author":"Wang","year":"2023"},{"key":"2026082813185013400_btag556-B43","doi-asserted-by":"crossref","first-page":"274","DOI":"10.1016\/j.immuni.2014.01.006","article-title":"Transcriptome-based network analysis reveals a spectrum model of human macrophage activation","volume":"40","author":"Xue","year":"2014","journal-title":"Immunity"},{"key":"2026082813185013400_btag556-B44","author":"Yoshida","year":"2017"},{"key":"2026082813185013400_btag556-B45","doi-asserted-by":"crossref","first-page":"201","DOI":"10.1038\/s41392-021-00572-w","article-title":"Small molecules in targeted cancer therapy: advances, challenges, and future perspectives","volume":"6","author":"Zhong","year":"2021","journal-title":"Signal Transduct Target Ther"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/advance-article-pdf\/doi\/10.1093\/bioinformatics\/btag556\/70100411\/btag556.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/42\/8\/btag556\/70100411\/btag556.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/42\/8\/btag556\/70100411\/btag556.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,28]],"date-time":"2026-08-28T17:19:02Z","timestamp":1787937542000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/doi\/10.1093\/bioinformatics\/btag556\/8742075"}},"subtitle":[],"editor":[{"given":"Jianlin","family":"Cheng","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,7,25]]},"references-count":45,"journal-issue":{"issue":"8","published-print":{"date-parts":[[2026,8,3]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btag556","relation":{},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"value":"1367-4803","type":"print"},{"value":"1367-4811","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,8]]},"published":{"date-parts":[[2026,7,25]]},"article-number":"btag556"}}