A generative model of identifying informative proteins from dynamic PPI networks

摘要: Informative proteins are the that play critical functional roles inside cells. They fundamental knowledge of translating bioinformatics into clinical practices. Many methods identifying informative biomarkers have been developed which heuristic and arbitrary, without considering dynamics characteristics biological processes. In this paper, we present a generative model by systematically analyzing topological variety dynamic protein-protein interaction networks (PPINs). model, common representation multiple PPINs is learned using deep feature generation based on original rebuilt reconstruction errors analyzed to locate proteins. Experiments were implemented data yeast cell cycles different prostate cancer stages. We analyze effectiveness comparing methods, ranking results also compared with from baseline methods. Our method able reveal members in progresses can be further studied testify possibilities for biomarker research.