One potential extension of the results and methodology seen in this study is to whole-genome sequencing efforts like the 1000 Genomes Project. One study design that has arisen in this context is to sequence many individuals at low coverage; say, 2–4×. The data from this kind of experiment are too sparse to directly and confidently determine most genotypes, but they can be called with high accuracy by applying the same kinds of models that are used for genotype imputation in GWAS (Li et al. 2011; The 1000 Genomes Project Consortium 2010). We expect that the approach of combining information across populations will help call low-frequency alleles in that setting, much as it helped impute low-frequency alleles in this study.
