Systems biology has become a major field of post-genomic bioinformatics research. A biological network containing various objects and their relationships is a fundamental way to represent a bio-system. A graph consisting of vertices and edges between these vertices is a natural data structure to represent biological networks. Substructure analysis of metabolic pathways by graph-based relational learning provides us biologically meaningful substructures for system-level understanding of organisms.
This chapter presents a graph representation of metabolic pathways to describe all features of metabolic pathways and describes the application of graph-based relational learning for structure analysis on metabolic pathways in both supervised and unsupervised scenarios. We show that the learned substructures can not only distinguish between two kinds of biological networks and generate hierarchical clusters for better understanding of them, but also have important biological meaning.