Modules in Biological Networks: Identification and Application

Introduction

The twentieth-century biology has been focused on individual cellular components and their functions. Despite the huge success of this approach, a discrete biological function can rarely be attributed to an individual molecule (Hartwell et al., 1999). It is increasingly clear that the cell can be understood as a complex network of interacting components (Barabasi & Oltvai, 2004). Unraveling the interactions between the components of a cell constitutes a major goal of the post-genomic era.

With recent advances in high-throughput experimental technologies, genomic data are now available for the reconstruction of large-scale biological networks, in which nodes are biological molecules (e.g. proteins, genes, metabolites, microRNAs, etc.) and edges are functional relationships among the molecules (e.g. protein interactions, genetic interactions, transcriptional regulations, protein modifications, metabolic reactions, etc.). Biological networks that have been studied include protein interaction networks constructed from protein-protein interaction data (Gavin et al., 2006; Ito et al., 2001; Krogan et al., 2006; Rual et al., 2005; Stelzl et al., 2005; Uetz et al., 2000), gene co-expression networks constructed from gene expression profiling data (Oldham et al., 2006; Stuart et al., 2003; van Noort et al., 2004), transcriptional regulation networks constructed from protein-DNA interaction data (Harbison et al., 2004; Lee et al., 2002), and metabolic networks constructed from bioreaction data (Duarte et al., 2007; Jeong et al., 2000). At a more abstract level, functional association networks have been used to represent integrated information from various types of functional association data (Franke et al., 2006; Jensen et al., 2009).

One of the most important properties of networks representing complex systems is modularity, i.e., the organization of nodes in clusters, with many edges connecting nodes of the same cluster and comparatively few edges connecting nodes of different clusters (Girvan & Newman, 2002). Indeed, modularity has been observed in protein interaction (Gavin et al., 2006), transcriptional regulation (Ihmels et al., 2002), and metabolic networks (Ravasz et al., 2002). Network-based module identification has captured the attention of a diverse group of scientists from various domains such as statistical physics, computer science, discrete mathematics, sociology, and computational biology (Fortunato, 2009). Although an ideal solution remains to be reached, the enormous effort of a large interdisciplinary community of scientists has generated a variety of approaches to this problem. In the biology community, besides network-based inference, modules can also be derived from existing knowledge on pathways and biological processes (Wang et al., 2008).

Modules, by definition, are sub-groups of elements (e.g. nodes and edges in the context of networks) that function in a semi-autonomous fashion and serve as building blocks of complex systems. The ability to decompose complex biological systems into modules allows the use of modules rather than individual genes as units in biological studies. Module-based analyses have several advantages over gene-based methods, including improved robustness against the inherent noise that exists in the sample population and increased sensitivity in identifying patterns that are too subtle to discern when considering individual genes (Chuang et al., 2007; Mootha et al., 2003). More importantly, module-based analyses can achieve a higher-level understanding of the design and organization of biological systems (Gavin et al., 2006; Segal et al., 2004).