Related Experiment Video
Updated: Aug 7, 2025
08:51
Author Spotlight: Integrated Multi-Omics Analysis for Unveiling Multicellular Immune Signatures in Clinical Heart Attack Cohorts
Published on: September 20, 2024
1.4K
Stabl: sparse and reliable biomarker discovery in predictive modeling of high-dimensional omic data
Julien Hédou1, Ivana Marić2, Grégoire Bellan3
1Department of Anesthesiology, Perioperative & Pain Medicine, Stanford University, Stanford, CA.
Research Square
|March 13, 2023
Summary
Stabl, a new machine learning framework, enhances biomarker discovery and clinical prediction by selecting sparse, reliable biomarkers. It improves upon existing methods for multi-omic data integration, aiding clinical translation.
Area of Science:
- Computational Biology
- Biostatistics
- Machine Learning
Background:
- High-content omic technologies and sparsity-promoting regularization methods (SRM) have advanced biomarker discovery.
- Translating computational biomarker findings into clinical applications is hindered by the challenge of selecting reliable candidates from complex multivariate models.
Approach:
- Propose Stabl, a machine learning framework unifying biomarker discovery with multivariate clinical outcome prediction.
- Stabl selects a sparse and reliable set of biomarkers, improving model interpretability and clinical utility.
Key Points:
- Stabl demonstrates superior biomarker sparsity and reliability compared to existing SRMs on synthetic and real-world clinical data.
- Achieves comparable predictive performance to other SRMs while enhancing biomarker selection rigor.
- Stabl effectively integrates multi-omic data (double- and triple-omics), identifying sparser, more reliable biomarkers than current fusion methods.
Conclusions:
- Stabl facilitates the biological interpretation and clinical translation of complex multi-omic predictive models.
- The framework offers a robust solution for selecting reliable biomarkers from high-dimensional omic data.
- Stabl's open-source availability supports wider adoption in biomarker research and clinical practice.
