Biological Relevance Detection via Network Dynamic Analysis

Most existing approaches for gene selection are based on evaluating the statistical relevance. However, there are remarkable discrepancies between statistical relevance and biological relevance. It is important to consider biological relevance for crucial genes identification. The task of detecting biological relevance presents two major challenges: first, how to define different types of measures to evaluate the biological relevance from multiple perspectives; and second, how to effectively integrate these measures to achieve better estimations. In this work, we propose to detect biological relevance by applying dynamics analysis using both biological networks and gene expression profiles from different phenotypes, and develop an effective probabilistic model to integrate various types of relevance measures in a unified form. Experimental results show the efficacy and potential of the proposed approach with promising findings.