CMStatistics 2017: Start Registration
View Submission - CMStatistics
Title: Personalized cancer-specific integrated network estimation Authors:  Kim-Anh Do - University of Texas MD Anderson Cancer Center (United States) [presenting]
Veerabhadran Baladandayuthapani - UT MD Anderson Cancer Center (United States)
Min Jin Ha - UT MD Anderson Cancer Center (United States)
Abstract: The aim is to propose personalized cancer-specific integrated network estimation (PRECISE), a general framework for integrating existing interaction databases, data-driven de novo causal structures, and upstream molecular profiling data to estimate cancer-specific integrated networks, infer patient-specific networks and elicit interpretable pathway-level signatures. We develop a Bayesian regression model for protein-protein interactions that integrates with known pathway annotations and protein-protein interactions. In the first step of PRECISE, a data-driven protein causal network is estimated and combined with the prior information. Other upstream molecular data are then integrated in three sequential steps to produce cancer-specific, patient-specific networks and pathway scores, which are subsequently used for tumor subtype classification and clinical outcome prediction. Using the pan-cancer functional proteomic data on 32 cancer types from The Cancer Genome Atlas, we demonstrate the utility of PRECISE in inferring commonalities and differences in network biology across tumor lineages and in using patient-specific pathway-based signatures for robust tumor stratification and prediction.