Detecting SNP Combinations Discriminating Human Populations From HapMap Data

The genomes of different human beings are similar. There are only a relatively small number of genetic differences between people. The genetic differences between people are very worthy of study. Researchers have proposed the fixation index
$F_{ST}$
measurement to find the single nucleotide polymorphisms (SNPs) which can reflect human population differences. However, most SNPs have interactions and they work together, which leads to the differences among human populations. The number of all possible
$m$
-locus combinations chosen from
$n$
SNPs grows exponentially. Most methods concern on 2-locus interactions. In this paper, we propose a novel method to find a new coordinate system under which the energy distributions of different populations are quite different. We select out candidate SNPs from
$n$
SNPs by using the information of the axes in the coordinate system. The number of candidate SNPs is small, thus SNP-SNP interactions can be searched efficiently. The method can also find interactions of more than two loci. These interactions should be able to reflect the evolution of human populations from another way. The numbers of SNP-SNP interactions are regarded as the differences between pairwise populations and a hierarchical clustering algorithm is used to construct the evolutionary tree. In the experiments, we apply the method to SNP data of four chromosomes separately and the trees constructed on these four chromosomes are highly consistent. Furthermore, the trees are also consistent with previous studies, which indicates that evolutionary information is well mined. The method provides a new insight to analyze the human population differences.