Feature ranking based on synergy networks to identify prognostic markers in DPT-1

Traditional epidemiologic methods test hypotheses focusing on individual risk factors for studying disease of interest. However, complex diseases are triggered and progress due to complicated interactions among both genetic and environmental risk factors. In this paper, we propose a network-based approach by integration of pairwise synergistic interactions to identify potential risk factors and their interactions in disease development. Specifically, we study immunologic and metabolic indices that may provide prognostic and diagnostic information regarding the development of Type-1 Diabetes (T1D) by analyzing measurements from oral glucose tolerance tests (OGTTs) and intravenous glucose tolerance tests (IVGTTs) in subjects with high risk from the Diabetes Prevention Trial-Type 1 (DPT-1) study. Performance comparison of our network-based method with individual factor based analysis demonstrates that the systematic analysis of all potential factors by considering their synergistic relationships help predict the development of clinical T1D better.