Information theoretic methods for detecting multiple loci associated with complex diseases

Gene-gene interactions play important roles in the etiology of complex multi-factorial diseases. With the advancements in genotyping technology, large genetic association studies based on hundreds of thousands of single-nucleotide polymorphisms are a popular option for the study of complex diseases. Association studies using locus by locus analyses till remains the primary method although the study of gene-gene interactions has become more common using regression based methods. However, regression based methods are computationally heavy and model complexity increases rapidly with the increase in number of loci and also with the number of possible allelic states at each locus. Information theoretic approaches offer many potent capabilities and advantages for the analyses of gene-gene interactions. In this paper, we develop and explore the effectiveness of two information theoretic metrics in identifying gene-gene interactions using extensive simulations on four different gene-gene interaction models. We propose a forward selection algorithm using the metrics and evaluate its performance using the rheumatoid arthritis dataset from Genetic Analysis Workshop-15. We demonstrate that our metrics are capable of analyzing a diverse range of epidemiological data sets containing evidences for gene-gene interactions.