The International HapMap 3 Consortium has expanded the HapMap resource by genotyping 1.6 million common SNPs and sequencing ten 100-kilobase regions in 692 individuals from 11 global populations. This integrated data set, called 'HapMap 3', includes both SNPs and copy number polymorphisms (CNPs). The study characterized population-specific differences among low-frequency variants, measured the improvement in imputation accuracy with a larger reference panel, and demonstrated the feasibility of imputing newly discovered CNPs and SNPs. The results highlight the need to characterize population-genetic parameters in each population and for each stratum of allele frequency, as lower-frequency variation is less shared across populations. The study also found that variants discovered through large-scale sequencing have longer haplotypes than more common variants, and that imputation can perform well for both CNPs and low-frequency SNPs, although success was partial and required specific conditions. The ultimate utility of such a reference panel will depend on the distribution of causal alleles across traits and the cost and accuracy of sequencing compared to genotyping followed by imputation.The International HapMap 3 Consortium has expanded the HapMap resource by genotyping 1.6 million common SNPs and sequencing ten 100-kilobase regions in 692 individuals from 11 global populations. This integrated data set, called 'HapMap 3', includes both SNPs and copy number polymorphisms (CNPs). The study characterized population-specific differences among low-frequency variants, measured the improvement in imputation accuracy with a larger reference panel, and demonstrated the feasibility of imputing newly discovered CNPs and SNPs. The results highlight the need to characterize population-genetic parameters in each population and for each stratum of allele frequency, as lower-frequency variation is less shared across populations. The study also found that variants discovered through large-scale sequencing have longer haplotypes than more common variants, and that imputation can perform well for both CNPs and low-frequency SNPs, although success was partial and required specific conditions. The ultimate utility of such a reference panel will depend on the distribution of causal alleles across traits and the cost and accuracy of sequencing compared to genotyping followed by imputation.