A flexible novel approach to learn epistasis based on ant colony optimization

Epistasis learning is becoming more and more important in genome-wide association (GWA) studies as single loci association test could only explain a small proportion of heredity of common human diseases. Many epistasis or gene-gene interaction learning methods have come out to try to solve the problem which is on the way of learning epistasis, which is computable difficult to search for high order epistatic interactions underlie common human disease and they have shown quite an ability to work on such a problem. But as single nucleotide polymorphism´s (SNP) dimensionality is growing larger and larger in current generated genome-wide association datasets, of which the number of samples that could be used to do GWA analysis is smaller relatively. How to solve the problem of epistasis learning caused by this dimensionality disaster is still a critical challenge for our genetic scientists. In this article, we propose a novel approach which could find a gene-gene interaction model consists of a flexible number of susceptible loci based on ant colony optimization (ACO) strategy and we conduct a lot of experiments on a wide range of simulated datasets and compare the outcome of our ACO method with some other epistasis learning methods like Bayesian combinational method (BayCom) and Multi beam search method (MBS), found that our ACO method is quite available and time efficient to solve the haystack problem to learn epistatic interactions and it may become a potential solution to search for complex association rules between susceptible SNP subset and common human disease in the future.