DNA sequencing cost of $133 per 1x sequencing (based on the Illumina Network cost of $4,000 for 30x sequencing of 50 samples or more, which scales linearly with lower coverage). We calculated the effective sample size of a sequencing-based GWAS as a function of average coverage, which determines the number of samples sequenced under a fixed budget (Online Methods). Under zero sample preparation cost and ignoring the benefit of imputation, the optimal study design involves sequencing a maximal number of samples at minimal coverage22,23. However, when sample preparation cost and imputation are taken into account, there exists an optimal number of samples to sequence for any budget. For a fixed budget of $300,000, the highest effective sample size (roughly equivalent to more than 4,600 typed individuals) is achieved at an average coverage of 0.1x (6,800 samples sequenced at $45 total cost per sample, r2=0.65) (Figure 4a). The optimal value of average coverage varies as function of sample preparation and sequencing costs, but we obtained qualitatively similar results for other cost assumptions (Supplementary Note). We note that a sequencing-based approach can attain a higher effective sample size than SNP arrays even when constraints on sample availability limit the space of available
