Fig. 1: Overview of the bio-primed LASSO framework for biomarker discovery.
From: Bio-primed machine learning to enhance discovery of relevant biomarkers
A Biomarker discovery analysis associates the dependency of the target gene with a genome-wide omic profile. Information from biological networks, such as PPI networks, is integrated into the LASSO regularization, bio-primed model. As a consequence, features that are linked to the target dependency are prioritized during the feature selection resulting in the discovery of relevant biomarkers. B Stepwise parameterization procedure first optimizes \(\lambda\) then second optimizes the Φ parameter. Using the optimized \(\lambda\) and Φ parameters a final LASSO model is fit, and the resulting coefficients can be inspected to prioritize biologically relevant biomarkers.
