Applying Enumerative, Spectral and Hybrid Graph Analyses to Biological Network Data

Abstract

Many biological data sets are expressible as complex networks of metabolic or genetic or protein structural information. As such they can be analysed and interpreted using enumerative graph methods for calculating their static and structural properties. We consider enumerative as well as spectral methods for analysing graphs of biological networks. We review computational algorithms for determining properties such as: path lengths; component cluster distribution; circuits and community structure as well as examining some heuristics and hybrid algorithms combining these approaches. We discuss some of the common properties they reveal for bio-network data. We apply these methods to some public domain biological network data and discuss the computational performance and scaling of these approaches to very large bio-network sizes. As well as considering hybrid combinations of these methods, we review which can best be applied to which size of bio-network data set.