Comparing multiple networks using the Co-expression Differential Network Analysis (CoDiNA)

Abstract

Biomedical sciences are increasingly recognising the relevance of gene co-expression-networks for analysing complex-systems, phenotypes or diseases. When the goal is investigating complex-phenotypes under varying conditions, it comes naturally to employ comparative network methods. While approaches for comparing two networks exist, this is not the case for multiple networks. Here we present a method for the systematic comparison of an unlimited number of networks: Co-expression Differential Network Analysis (CoDiNA) for detecting links and nodes that are common, specific or different to the networks. Applying CoDiNA to a neurogenesis study identified genes for neuron differentiation. Experimentally overexpressing one candidate resulted in significant disturbance in the underlying neurogenesis' gene regulatory network. We compared data from adults and children with active tuberculosis to test for signatures of HIV. We also identified common and distinct network features for particular cancer types with CoDiNA. These studies show that CoDiNA successfully detects genes associated with the diseases.