SEE: a novel multi-objective evolutionary algorithm for identifying SNP epistasis in genome-wide association studies

Abstract

Although genome-wide association studies play an increasingly important role in identifying causes of complex diseases, detecting SNP epistasis in these studies is a computational challenge. The existing methods are usually based on a single-correlation model between SNP combinations and phenotype and their performance is often unsatisfactory. The highest average power of the existing methods is 0.58 on DME models and 0.97 on DNME models. The highest average F-measure of the existing methods is 0.44 on DME models and 0.90 on DNME models. The lowest average computation time (second) of the existing methods is 2.12 on DME models and 2.09 on DNME models. In this work, a novel multi-objective evolutionary algorithm named SEE is presented for identifying SNP epistasis. In SEE, eight evolution objectives are successfully integrated to measure the association between SNP combinations and phenotype. SEE uses a novel evolutionary strategy based on sort, exploration and exploitation. SEE was compared with other existing methods using 72 simulated datasets. The average power of SEE is 0.71 with DME models and 0.99 with DNME models. The average F-measure of SEE is 0.68 with DME models and 0.99 with DNME models. The average computation time of SEE is 0.21 with DME models and 0.40 with DNME models. It is indicated that SEE outperforms other algorithms in both F-measure and computation time. It was then utilized to analyze real data obtained from the Wellcome Trust Case Control Consortium. Availability and Implementation: SEE is freely available at https://github.com/sunliyan0000/SEE.