<link rel="stylesheet" href="styles.f3b1fba60ec7970c.css">

Discovery of sparse, reliable omic biomarkers with Stabl

dc.creatorHédou, Julien
dc.creatorMarić, Ivana
dc.creatorBellan, Grégoire
dc.creatorEinhaus, Jakob
dc.creatorGaudillière, Brice
dc.date2023-10-12
dc.date.accessioned2026-11-02T10:11:38Z
dc.description&lt;p&gt;&lt;span&gt;Adoption of high-content omic technologies in clinical studies, coupled with computational &lt;/span&gt;&lt;span&gt;methods, have yielded an abundance of candidate biomarkers. However, translating such find&lt;/span&gt;&lt;span&gt;ings into bona fide clinical biomarkers remains challenging.&lt;/span&gt; &lt;span&gt;To facilitate this process, we &lt;/span&gt;&lt;span&gt;introduce Stabl, a general machine learning framework that identifies a sparse, reliable set &lt;/span&gt;&lt;span&gt;of biomarkers by integrating noise injection and a data-driven signal-to-noise threshold into &lt;/span&gt;&lt;span&gt;multivariable predictive modeling.&lt;/span&gt; &lt;span&gt;Evaluation of Stabl on synthetic datasets and five inde&lt;/span&gt;&lt;span&gt;pendent clinical studies demonstrates improved biomarker sparsity and reliability compared to &lt;/span&gt;&lt;span&gt;commonly used sparsity-promoting regularization methods while maintaining predictive per&lt;/span&gt;&lt;span&gt;formance; it distills datasets containing 1,400 to 35,000 features down to 4 to 34 candidate &lt;/span&gt;&lt;span&gt;biomarkers. Stabl extends to multi-omic integration tasks, enabling biological interpretation of &lt;/span&gt;&lt;span&gt;complex predictive models, as it hones in on a shortlist of proteomic, metabolomic, and cyto&lt;/span&gt;&lt;span&gt;metric events predicting labor onset, microbial biomarkers of preterm birth, and a pre-operative &lt;/span&gt;&lt;span&gt;immune signature of post-surgical infections.&lt;/span&gt;&lt;/p&gt;
dc.description&lt;p&gt;Funding provided by: Stanford University&lt;br&gt;Crossref Funder Registry ID: http://dx.doi.org/10.13039/100005492&lt;br&gt;Award Number: &lt;/p&gt;
dc.identifierhttps://doi.org/10.5281/zenodo.8406758
dc.identifieroai:zenodo.org:8406758
dc.identifier.urihttps://library.dare.co.zw/handle/123456789/171
dc.publisherZenodo
dc.relationhttps://doi.org/10.1038/s41586-022-04410-z
dc.relationhttps://doi.org/10.1016/j.patter.2022.100655
dc.relationhttps://doi.org/10.1016/j.xcrm.2022.100680
dc.relationhttps://doi.org/10.1016/j.xcrm.2021.100287
dc.relationhttps://doi.org/10.1101/2023.03.07.23286920
dc.relationhttps://doi.org/10.5061/dryad.stqjq2c7d
dc.relationhttps://zenodo.org/communities/dryad/
dc.relationhttps://doi.org/10.5281/zenodo.8406757
dc.rightsinfo:eu-repo/semantics/openAccess
dc.rightsBSD 3-Clause Clear License
dc.rightshttp://labs.metacarta.com/license-explanation.html#license
dc.subjectBiomarkers
dc.subjectMachine learning
dc.subjectfeature selection
dc.titleDiscovery of sparse, reliable omic biomarkers with Stabl
dc.typeinfo:eu-repo/semantics/other

Files

Collections