The coverage of human genomic variations is known to substantially affect the success of genome-wide association studies. We therefore assessed the SNP coverage in the HapMap database for a total of 1,304 subjects from the Japanese population by combining resequencing and high-density genotyping approaches. First, we resequenced 48 Japanese subjects in 86 genes (572 kb in total), and we then genotyped the subset of tag SNPs and also imputed genotypes for all of the detected SNPs in an additional panel of 1,256 subjects. Subsequently, we genotyped 555,352 tag SNPs selected from the HapMap in 72 Japanese subjects (from the panel of 1,256 subjects) and further imputed genotypes for all SNPs currently included in the HapMap. Of 738 common genic SNPs (1.3 per kb) that we detected by resequencing, 58% had already been genotyped in the HapMap, and 31% were not genotyped but had a proxy SNP in the HapMap with a linkage disequilibrium coefficient r (2) > or = 0.8, whereas 11% were not represented in the current HapMap database. Thus, the HapMap coverage appears to be high although not thorough for SNPs in the Japanese population as compared to its coverage reported in Caucasians, and this needs to be considered when we interpret association results.