Identification of Minimum Redundancy Tagging SNPs via Gibbs Sampling

Single nucleotide polymorphisms (SNPs) are genetic changes that can occur within a DNA sequence. Due to the high frequency of SNPs in the human genome, it is desirable to select a small set of SNPs (tagging SNPs) that can be used to represent the majority of SNPs. We propose a Gibbs sampling approach to find a small set of SNPs with minimum redundancy for tagging purposes. Preclustering is added in the basic Gibbs sampling procedure to avoid the disturbance caused by local optima. We also propose two general purpose correlation measures that are able to accommodate SNPs with three or more alleles. Our experimental results show that Gibbs sampling process converges faster and finds better optimum if pre-clustering is conducted before the sampling process. While our tagging process is not guided by any prediction algorithm, we are able to obtain comparable results as the SNP prediction guided algorithm SVM/STSA [1] while requiring much less time.