is genotyped then the genotype at the others can be probabilistically inferred via imputation. The ability to infer surrounding genotypes based on LD patterns means that we can now cost-efficiently scan the genome with a much smaller subset of markers than previously needed (e.g., in the range of 370,000–600,000) and impute the remaining commonly occurring markers across the genome. Whereas early attempts to impute variation were restricted to common SNPs (>5% minor allele frequency), the landmark 1000 Genomes Project (http://www.1000genomes.org/) used a similar strategy to identify less common SNPs (≤1%). Our growing knowledge of genetic variants across the genome, coupled with exponential decreases in costs of high density GWAS arrays (>1 million SNPs at <100 USD per participant currently), now allows investigators to interrogate over 10 million polymorphisms in relation to outcomes.
