Akey et al. and Barreiro et al. used Weir and Cockerham��s estimate, an unbiased estimate of Fst. Casto et al. used four measures: ��, the difference in allele frequency CITCO between two groups; integrated haplotype score, which characterizes the lengths of the haplotypes surrounding each allele of a SNP ; latitude/longitude correlation, which describes how closely changes in a SNP��s allele frequency follow geographical coordinates; and Fst, which shows variation in allele frequency among populations. Park et al. used the Nearest Shrunken Centroid Method, which was originally designed for clustering of microarray data. NSCM has been proposed for solving the classification problem with a large number of features and it was also applied to the analysis of population differentiation in SNPs via Hapmap data. Han et al. modified Fst for use with allele frequency data with unbalanced sample sizes. In order to investigate PD of DR genes, we first compared four measures for assessing population differentiation: the chi-square test, the ANOVA F-test, Fst, and NSCM. Fst showed high sensitivity with stable specificity among varying sample sizes; thus, we selected Fst for determining population differentiation.We then divided DR genes from PharmGKB into two groups based on the degree of population differentiation as assessed by Fst: genes with high a level of differentiation and genes with a low level of differentiation. Finally, we conducted a gene ontology analysis and pathway analysis. Several studies have investigated PD associated with individual drugs. In the present study, we systematically studied PD of drug-related genes by simultaneously considering all reported DR genes. This integrative approach may help clarify the inconsistent genetic features of drug response associated with PD. Furthermore, our findings will improve the study and prediction of drug responses that differ among populations due to genetic stratification. To investigate the biological differences between the HD and LD gene groups, we performed a GO analysis and a pathway analysis using the Database for Annotation, Visualization and Integrated Discovery v6.7 functional annotation tool. Annotated genes from each group were used as the input, while a list of whole genes in DAVID with at least one annotation in the analyzing categories was used as the background. For the GO analysis, the following three categories were selected: biological process, molecular function, and cellular component. For the pathway analysis, the Kyoto Encyclopedia of Genes and Genomes pathway was used. Additional GO and pathway analyses were performed in a similar manner in order to compare genes in the HD gene group to those in the DR gene group. In this case, the DR HD gene group was used as the input for analysis, and the DR gene group was used as the background. To correct for multiple tests, we used the hypergeometric test from Benjamini-Hochberg��s method. Fold enrichments, defined as the ratios of proportions between the input and background, were calculated for each term. Terms with Benjamini-Hochberg��s q-values of 0.05 or lower were considered significant. PD is important for understanding differences in drug responses among populations. However, PD often refers to the distance between two different subpopulations; therefore, several studies have investigated approaches for averaging the PD of each SNP. For CP 93129 dihydrochloride instance, the impact of SNP ascertainment on estimating the distance between subpopulations has already been reported.