Elements of Computational Systems Biology

7

RECONSTRUCTION OF BIOLOGICAL NETWORKS BY SUPERVISED MACHINE LEARNING APPROACHES

Jean-Philippe Vert

Mines ParisTech - Institut Curie, Paris, France

7.1 INTRODUCTION

In this review chapter, we focus on the problem of reconstructing the structure of largescale biological networks. By biological networks, we mean graphs whose vertices are all or a subset of the genes and proteins encoded in a given organism of interest, and whose edges, either directed or undirected, represent various biological properties. As running examples, we consider the three following graphs, although the methods presented below may be applied to other biological networks as well.

• Protein-protein interaction (PPI) network. This is an undirected graph with no self-loop, which contains all proteins encoded by an organism as vertices. Two proteins are connected by an edge if they can physically interact
• Gene regulatory network. This is a directed graph that contains all genes of an organism as vertices. Among the genes, some called transcription factors (TFs) regulate the expression of other genes through binding to the DNA. The edges of the graph connect TFs to the genes they regulate. Self-loops are possible. if a TF regulates itself. Moreover, each edge may in principle be labeled to indicate whether the regulation is a positive (activation) or negative (inhibition) regulation.
• Metabolic network. This graph contains only a subset of the genes as vertices, namely, those coding for enzymes. Enzymes are proteins ...