Probabilistic graphical modeling in systems biology: A framework for integrative approaches

Abstract

Systems biology may be defined as a discipline aiming at integrating various sources of heterogeneous data, with the objective to describe and predict the function of biological systems. The purpose is to cross many (possibly weak) evidences from several data types that describe different biological features of genes or proteins. Probabilistic graphical models offer an appealing framework for this objective. Through the thorough review of five selected examples, this chapter highlights how probabilistic graphical models can contribute to build the bridge between biology and computational modeling. In this methodological framework, the five cases illustrate three features of these models, which we discuss: flexibility, scalability and ability to combine heterogeneous sources of data. The applications covered address genetic association studies, identification of protein–protein interactions, identification of the target genes of transcription factors, inference of causal phenotype networks and protein function prediction.