Prediction of Disease-associated Single Nucleotide Polymorphisms Using Virtual Genomes Constructed from a Public Haplotype Database

Abstract

Simultaneous dealing of hundreds of thousands of single nucleotide polymorphisms (SNPs) in genome-wide association studies is laborious. The aim of our study is to develop a method to decrease the number of candidate SNPs prior to the genotyping of study subjects. We created virtual genotype data on case and control subjects from data of the International HapMap Project by using haplotype-based simulation method. We repeated virtual case-control association studies and selected candidate SNPs. We applied the selected SNPs to previously published genetic case-control studies. Sensitivity to detect association with causative genes using our method was compared to the original studies and results using tag SNPs. We found a discrete distribution pattern of SNPs, which was able to produce significant results in case-control association studies. The number of candidate SNPs that we selected was 24.7% of the number of the original set of SNPs. We applied this method to previously published genetic case-control studies and found that the use of candidate SNPs improved the sensitivity for detecting significant alleles, both compared to the original studies and to the use of tag SNPs. The results were not affected by the difference of the diseases and genes involved. Our simulation-based approach has advantages of reducing costs and improving performance to detect significant alleles. This method can be used without considering the specific disease and genes involved.