SNP specific extraction and analysis using shrunken dissimilarity measure

The identification of genetic variations associated with common disease now plays an important role in human genetics research. In this paper, a new nearest shrunken centroid method was performed to select relevant single nucleotide polymorphisms (SNPs) on a WTCCC Coronary Artery Disease data. This method can succinctly characterize each class (case and control) by shrinking each centroid with respect to the overall centroid by a certain threshold. A relatively high average accuracy of 87% among all 22 chromosomes can be obtained. There are 221 out of 490032 SNPs selected using the proposed shrunken centroid method under a 10-fold cross validation setting. The average number of SNPs being selected is around 10 for each of the chromosome. Comparisons with other shrunken centroid methods were performed, results showed that the performance of the proposed method in terms of accuracy and numbers of selected SNPs is better than others. All computational results show that the proposed shrunken centroid method is a suitable and useful tool to select relevant SNPs with genetic variations in a genome-wide association disease study.