Multiple Sclerosis Biomarker Discovery via Bayesian Feature Selection

Recent work proposes a hierarchical Bayesian framework for feature selection, where a prior describes the identity of each feature set and the underlying distribution parameters. Assuming jointly Gaussian features, a posterior is found in closed form, and an approximation is presented to develop fast suboptimal algorithms. Applying this method to multiple sclerosis data we find highly ranked genes and pathways suggested to be involved in multiple sclerosis.