Discovering Disease Associated Gene-Gene Interactions: A Two SNP Interaction Analysis Framework

Most common diseases have a heritable component that is influenced by mutations on multiple loci, and by interactions between loci and with the environment. However, traditional genetic analysis techniques have focused on single locus effects. This is mostly due to the polynomial increase in computational capacity needed to attempt multi-loci interaction analyses, and the anticipated loss of power due to multiple testing. In this dissertation, a framework for performing a complete two single nucleotide polymorphism (SNP) interaction analysis of high dimensionality genome wide association scans (GWAS) is presented. The implementation of the framework utilizes diverse distributed computational resources to overcome the bottlenecks of each resource, harvesting enough capacity to analyze any of the GWAS in existence today within a reasonable time frame. Algorithmic approaches are proposed to improve the efficiency of the framework and improve its computational performance so that a brute force attack on the problem can be performed. The data is encoded in binary using a lossless algorithm that significantly reduces its size. Computationally efficient data mining measures for the Omnibus and Epistatic interaction effects are proposed and compared to traditional statistical techniques. An algorithm is proposed that optimizes the analyses of multiple response variables within the same GWAS. GenMSA, a multiple sclerosis (MS) dataset, is analyzed using the proposed framework with top results tested for replication using ANZgene, an independent MS, dataset. Some of the top results replicated, implicating SNPs in a region of known association to MS providing evidence to the validity of the proposed framework. Top results are further examined through a proposed approach that enables drilling into these results and studying correlation coefficient between each of the genotype combinations of the SNP and the signal level to each of the main and epistatic effects.