An improved GA for identifying susceptibility genes in the presence of epistasis

Identifying the epistasis models between single nucleotide polymorphisms (SNPs) in several genes can explain the susceptibility to diseases. The statistical methods have been used to identify the significant epistasis models according to the related statistical values, including odds ratio (OR), chi-square test (χ²), p-value, etc. However, the high calculations limit the statistic to identify the high-order epistasis. In this study, we proposed an lsGA algorithm, genetic algorithm based on local search algorithm, to identify the significant epistasis model amongst the large SNP combinations. Two disease models were used to simulate the large data sets considering the minor allele frequency (MAF), number of SNP, and number of sample. The 3-order epistasis models were identified by chi-square test (χ²) for evaluating the significance (P-value < 0.05). lsGA was compared with GA to analyze the improvement in the search abilities, and results showed that lsGA provided higher chi-square test values than that of GA.