Analysis of SNP Interaction Combinations to Determine Breast Cancer Risk with PSO

Abstract

Many association studies analyze the genotype frequencies of case and control data to predict susceptibility to diseases and cancers. An increasing number of studies has shown that the risk of getting diseases and cancers is associated with the co-occurrence of some contain single nucleotide polymorphisms (SNPs). Determining the disease-causing SNPs has become an important objective. In order to study the SNP-SNP interaction in breast cancer, we used a particle swarm optimization (PSO) algorithm to compute the difference between the control and case data and performed a feature selection from different SNP combinations with their corresponding genotypes. The best combination of SNP-SNP interactions is the maximal difference of co-occurrences between the control and case groups. In this study, we explored the SNP interaction of 19 SNPs in 372 controls and 398 cases of breast cancer association using simulated SNP data of breast cancers. The odds ratio (OR) were used to evaluate the breast cancer risk in terms of the best combination of SNP-SNP interactions. Compared to their corresponding non-SNP combinations, the estimated OR of the best predicted SNP combination with genotypes for breast cancer is significantly greater than 1 (about 1.771 and 2.417; confidence interval (CI): 1.223-4.371; p <; 0.05-0.001) for specific SNP combinations of two to five SNPs. The SNP interaction associated with a high risk of breast cancer could be successfully predicted using the proposed PSO method.