CoDiNA: an RPackage for Co-expression Differential Network Analysis in n Dimensions
share
Biological and Medical science is increasingly acknowledging the use of gene co-expression networks for the analysis of complex systems or diseases. In many studies, the goal is not only to describe a network, but to investigate how this network is changing under different conditions, with respect to certain diseases, or between different species. While methods for comparing two networks exist, this is not the case for comparing multiple networks, although this is a common aim of many studies. Moreover, much of the power of network analyses is lost when restricting a comparison to solely determining presence/absence of nodes and links. Here we present a method for the comparison of an unlimited number of networks: Co-expression Differential Network Analysis (CoDiNA). Our method distinguishes between links that are common to all networks, links that are specific to only one of the compared networks, and links that are different in that their sign changes between networks. Importantly, we developed a statistical framework to normalize between these different categories of common or changed network links. Our approach also allows us to categorize nodes as dominated by common, specific, or differentiated links. We demonstrate the usage of our new method by presenting the analysis of a dataset consisting of different types of cancers. We revealed common, specific, and differentiated links between cancer types and show that each category of links is enriched for genes with distinct Gene Ontology groups. This suggests that our method can detect functional differences. CoDiNA has been implemented in an R package that is available from CRAN.
READ FULL TEXT