Integrative approaches for finding modular structure in biological networks

Abstract

A central goal of systems biology is to elucidate the structural and functional architecture of the cell. To this end, large and complex networks of molecular interactions are being rapidly generated for humans and model organisms. A recent focus of bioinformatics research has been to integrate these networks with each other and with diverse molecular profiles to identify sets of molecules and interactions that participate in a common biological function - that is, 'modules'. Here, we classify such integrative approaches into four broad categories, describe their bioinformatic principles and review their applications.

Figure 1. Identifying ‘network hotspots’

(A) Schematic representation of active modules inferred through integration of biological networks and cellular state profiles. (B) Common procedural workflow involved in active module identification. Panel C shows regulatory interactions of clear renal cell carcinoma cancer genes that were identified by integrating mutation, copy number and mRNA expression data, with pathway information catalogued in public databases using the PARADIGM algorithm. In particular the network identified roles for chromatin remodelers in this cancer. Each gene is depicted as a set of concentric rings representing various levels of biological information, and where each ‘spoke’ in a ring pertains to a single patient sample. From periphery inwards, the rings indicate ‘PARADIGM’-inferred levels of gene activity, mRNA expression levels, mutational abundance and correlation of gene expression or activity to mutation events. Reproduced with permission from Nature 2013.

Figure 2. Differential analysis of molecular networks across conditions

(A) Schematic representation of a differential mapping approach to identify dynamically rewired modules. Molecular networks are assembled under multiple static conditions and these static networks are subtracted across a pair of conditions to deduce differentially enriched interactions and modules. As illustrated, differential genetic modules may be further projected on protein-protein interactions to display functional (genetic) interrelations within and among protein complexes. (B) Differential rewiring of the Serine/threonine MAP kinase SLT2 protein-protein interactions before and after methyl methanesulfonate -induced DNA damage in yeast. Red and green edges indicate positive and negative differential interactions (i.e., corresponding pairwise-deletion phenotypes associated with these interactions are deemed favorable or detrimental for survival respectively). Adapted from Molecular systems biology (2012) (C) Differential genetic interactions can co-cluster between protein complexes (pathways). The network shows cross-pathway genetic interactions bridging distinct protein complexes (pathways) functioning in DNA damage response in S. cerevisiae Adapted from Science.

Figure 3. Integrating networks across interaction types

(A) Schematic view of composite functional modules identified through computational integration of genetic and physical networks. (B) A hierarchical module map extracted from network information obtained from ref using the Cytoscape based application tool, Pangia illustrating intra-modular and inter-modular relationships in a joint network of protein-protein interactions and genetic interactions in yeast. Modules are determined based on physical and genetic interaction densities. Functional modules are represented as boxes, while edges between boxes represent the density of inter-module genetic interactions, i.e. connecting genes across the two modules (C) Magnified internal view of four network modules revealing their physical interactions and distinct ‘within-pathway’ and ‘between-pathway’ genetic interactions, where blue and red edges symbolize protein-protein and genetic interactions respectively. Networks were produced using data from 130 and the cytoscape tool.

Figure 4. Identification of conserved functional modules by integration of data across multiple species

(A) Functional linkage networks are assembled from multiple lines of evidence (e.g., protein-protein and genetic interactions, gene expression, protein localization, phenotype, and sequence data) and integrated with differential gene-expression profiles, in this example derived from human and mouse tissues (stem cell and differentiated cells). Candidate seed genes (red) are defined as differentially expressed othologues). The functional neighborhood (yellow) of each seed gene is marked by genes whose path confidence (the product of linkage weights along the path) from the seed gene meets a specified threshold. (B) A search for modules seeks densely connected subnetworks of genes sharing similar patterns of expression in both species. (C) In this search, subnetworks are grown simultaneously in both species starting from the seed genes (red) and expanded through iterative addition of genes satisfying both of two criteria: first, the genes must be in the same functional neighborhood, and second, the genes must maximize a differential expression activity score. Differentially expressed genes are colored green (up) or red (down)