A novel method for identifying SNP disease association based on maximal information coefficient
Abstract
To improve single-nucleotide polymorphism (SNP) association studies, we developed a method referred to as maximal information coefficient (MIC)-based SNP searching (MICSNPs) by employing a novel statistical approach known as the MIC to identify SNP disease associations. MIC values varied with minor allele frequencies of SNPs and the odds ratios for disease. We used a Monte Carlo-based permutation test to eliminate the effects of fluctuating MIC values and included a sliding-window-based binary search whose time-cost was 0.58% that of a sequential search to save time. The experiments examining both simulation and actual data demonstrated that our method is computationally and statistically feasible after reducing the resampling count to 4 times the number of markers and applying a sliding-window-based binary search to the method. We found that our method outperforms existing approaches.
To improve single-nucleotide polymorphism (SNP) association studies, we developed a method referred to as maximal information coefficient (MIC)-based SNP searching (MICSNPs) by employing a novel statistical approach known as the MIC to identify SNP disease associations. MIC values varied with minor allele frequencies of SNPs and the odds ratios for disease. We used a Monte Carlo-based permutation test to eliminate the effects of fluctuating MIC values and included a sliding-window-based binary search whose time-cost was 0.58% that of a sequential search to save time. The experiments examining both simulation and actual data demonstrated that our method is computationally and statistically feasible after reducing the resampling count to 4 times the number of markers and applying a sliding-window-based binary search to the method. We found that our method outperforms existing approaches.