A striking finding is that the effective sample size achieved using sequencing-based GWAS with current costs11 is more than six times higher than SNP-array genotyping at $400/sample, corresponding to a large increase in power (Figure 4b, Supplementary Note, Supplementary Figure 8). Only if SNP array typing is less than $70 per sample, or if sample preparation and sequencing costs are much higher (e.g. greater than $120 per sample for sample preparation or $1,000 for 1x sequencing) does sequencing-based GWAS lose its advantage in terms of statistical power to associate variants. If sequencing technology—both in the efficiency of library preparation and the cost of sequencing—continues to improve more quickly than genotyping technology, the advantage of sequencing-based GWAS will increase. We note that a critical ingredient for attaining high accuracy at ultra low-coverage is the availability of large panels of reference haplotypes. As additional reference haplotypes over larger numbers of SNPs become available from 1000 Genomes Project and other projects, we expect the accuracy attained by ultra low-coverage sequencing to further increase.
