Ant colony optimization with an automatic adjustment mechanism for detecting epistatic interactions.

Single Nucleotide polymorphisms (SNPs) are usually used as biomarkers for research and analysis of genome-wide association study (GWAS). Moreover, the epistatic interaction of SNPs is an important factor in determining the susceptibility of individuals to complex diseases. Nowadays, the detection of epistatic interactions not only attracts attention of many researchers but also brings new challenges. It is of great significance to mine epistatic interactions from large-scale data for the combinatorial explosion problem of loci. Hence, it is necessary to improve an efficient algorithm for solving the problem. In this article, a novel ant colony optimization based on automatic adjustment mechanism (AA-ACO) is proposed. The mechanism automatically adjusts the behaviour of artificial ants according to the real-time feedback information so that the algorithm can run at its best. This study also compares AA-ACO with ACO, AntEpiSeeker, AntMiner, MACOED and epiACO in a set of simulated data sets and a real genome-wide data. As shown by the experimental results, the proposed algorithm is superior to the other algorithms.