Inferring weighted and directed gene interaction networks from gene expression data using the phi-mixing coefficient

In this paper, we present a new algorithm for reverse-engineering gene interaction networks (GINs) from expression data, using the so-called phi-mixing coefficient between two random variables. Unlike existing methods, the GINs constructed using the algorithm presented here have edges that are both directed and weighted. The GIN constructed is, in a very specific sense, a minimal network that is compatible with the data. Several GINs have been constructed for various data sets in lung cancer, ovarian cancer and melanoma. Lung cancer and melanoma networks have been validated by comparing their predictions against the output of ChIP-seq data. The neighbors of three transcription factors (ASCL1, PPARG and NKX2-1) in lung cancer, and one transcription factor SOX10 in melanoma, are significantly enriched with ChIP-seq genes compared to pure chance.