Combinatorial algorithms for discovering subnetwork markers for cancer subtypes
In traditional approaches for discovering gene expression biomarkers, each gene is first ranked on how well it can differentiate between normal and tumor samples or samples from different subtypes of a cancer. The top genes are then selected based on a significance threshold. These approaches offer limited insights into the interactions among gene markers. However, these interactions are essential to understanding underlying mechanisms of diseases. Therefore, recent research efforts on discovering gene markers integrate knowledge of molecular interactions such as protein-protein interactions (PPIs) with genome-wide expression profiles. In this project, I consider novel combinatorial algorithms to search for optimal connected subnetworks from PPI networks that predict different subtypes of colon cancer.