CChi: An efficient cloud epistasis test model in human genome wide association studies

Abstract

Due to the vast amounts of SNPs and huge search space, how to decrease the total computation costs is a challenge in genome wide association studies (GWAS). Triggered by this problem, we develop an effective and efficient algorithm for epistasis detection in GWAS. We propose a cloud-based algorithm using chi-square test, denoted as CChi. CChi adopts a pruning strategy by utilizing an upper bound to prune amounts of unnecessary SNP pairs, and is implemented under Google's MapReduce framework. A best-fit model is proposed by us to distribute SNP pairs to each reducer. Extensive experimental results demonstrate that CChi is practically and computationally efficient.