A Mutual Information-based method to select informative pairs of variables in case-control genetic association studies to improve the power of detecting interaction between genetic variants.
We propose a novel procedure for tagging Single Nucleotide Polymorphisms (SNPs), called EpiTag, to deal with interaction detection in Genome-Wide Association Studies. The aim of our method is to select a set of tag-SNPs that optimally represents the whole set of pairs of SNPs whereas usual approaches are univariate. The linkage between two pairs of SNPs is measured by the Normalized Mutual Information. The proposed algorithm is assessed considering the power of interaction detection compared to a no-tagging strategy and a usual one-dimensional tagging procedure, both on simulated and real genotype structures. EpiTag demonstrates good power performances along with various signal strengths or data sizes w.r.t the competing methods.