Inferring gene-gene interactions and functional modules using sparse canonical correlation analysis
Networks pervade many disciplines of science for analyzing complex systems with interacting components. In particular, this concept is commonly used to model interactions between genes and identify closely associated genes forming functional modules. In this paper, we focus on gene group interactions and infer these interactions using appropriate partial correlations between genes, that is, the conditional dependencies between genes after removing the influences of a set of other functionally related genes. We introduce a new method for estimating group interactions using sparse canonical correlation analysis (SCCA) coupled with repeated random partition and subsampling of the gene expression data set. By considering different subsets of genes and ways of grouping them, our interaction measure can be viewed as an aggregated estimate of partial correlations of different orders. Our approach is unique in evaluating conditional dependencies when the correct dependent sets are unknown or only partially known. As a result, a gene network can be constructed using the interaction measures as edge weights and gene functional groups can be inferred as tightly connected communities from the network. Comparisons with several popular approaches using simulated and real data show our procedure improves both the statistical significance and biological interpretability of the results. In addition to achieving considerably lower false positive rates, our procedure shows better performance in detecting important biological pathways.