Predictive modeling and reliable biomarker discovery in clinical omics studies
Predictive modeling and reliable biomarker discovery in clinical omics studies
High-content omic technologies coupled machine learning methods have transformed the biomarker discovery process. However, the translation of computational results into scalable clinical biomarkers remains challenging. A rate-limiting step is the rigorous selection of reliable biomarker candidates among a host of biological features. Drawing examples from real-world clinical omics studies, I will introduce Stabl, a general machine learning framework that identifies a sparse, reliable set of biomarkers by integrating noise injection and a data-driven signal-to-noise threshold into multivariable predictive modeling.
Julien Hédou
CEO, SurgeCare
Julien Hédou studied Engineering at Ecole Polytechnique (Paris) before completing his MS in Biomedical Informatics degree from Stanford University. Julien has been working on developing the machine learning pipelines of the Gaudillière lab that integrates high parameter mass cytometry and proteomics using sparse machine learning approaches to identify biologically plausible and reliable predictive biomarkers, focusing on several clinical scenarios: 1) immune mechanisms of surgical recovery and complications, 2) feto-maternal health outcomes, including preterm birth, preeclampsia and endometriosis, 3) immune dysfunction and outcomes prediction in patients with acute ischemic stroke. Julien Hédou is now CEO of SurgeCare, a life science company that commercializes, PreCyte®, for predicting the risk of postoperative complications, and its laboratory services, SurgeLab™, for acquiring individualized immune signatures and analyzing multiomic datasets.
Predictive modeling and reliable biomarker discovery in clinical omics studies
A presentation by Julien Hédou
More webinars
WebinarPathways to Digital Health: AI and Omics in Rheumatoid Arthritis
Explore how groundbreaking proteomic research is transforming our understanding of rheumatoid arthritis (RA). In this on-demand webinar, Allan Stensballe, PhD, shares new insights into the molecular landscape of RA-affected synovial tissue, revealing how autoantibodies and protein signatures may hold the key to more precise personalized therapies.
WebinarUsing Antibody Profiling to Identify Novel Diagnostic Biomarkers
Current cancer screening methods often lead to false positives, false negatives and invasive biopsies that lack prognostic insights. Emerging research suggests that cancer-specific IgM and IgG antibodies – produced by B cells upon recognizing malignant cells – could serve as stable, easily measurable blood biomarkers for detecting and monitoring high-incidence cancers like melanoma and breast, prostate, bowel, lung and pancreatic cancer. This approach has the potential to improve early diagnosis, reduce uncertainty and enhance treatment planning.
WebinarLinking the plasma proteome to echocardiographic parameters in patients with chronic heart failure
Recent advancements in proteomic profiling have unlocked new possibilities for understanding the complex mechanisms that drive heart failure. Measuring thousands of proteins simultaneously makes it possible to capture a comprehensive overview of the patient's health state and investigate underlying disease progression at a subclinical level.