Probable Biomarker Identification Using Recursive Feature Extraction and Network Analysis

Biomarkers have tremendous potential in different phases of treatment such as risk assessment, screening/detection, diagnosis and patient's response prediction. In this paper, we present an approach for development of a generic tool for an end to end analysis of expression data to identify the probable biomarkers. We follow machine learning as well as network analysis approaches in parallel. We use statistical techniques as preliminaries for quality analysis, followed by the feature (gene) selection approach. For network analysis techniques we use measures such as eigen centrality, closeness centrality and betweenness centrality to filter the most influential mutated genes which act as biomarkers.