genotypes at all polymorphic sites indentified in the European samples of the 1000 Genomes Project, using sequencing data together with the 762 haplotypes inferred from the European samples of the 1000 Genomes Project phase 1 (Online Methods), and quantified accuracy by comparing imputed calls with Illumina array genotyping calls (Online Methods). To remove effects of high coverage at or near exons we removed data at all SNPs covered at more than 4x (Supplementary Figure 2). At 0.24x coverage we observe an average r2=0.71 (s.d. 0.15) to the genotype calls assayed by genome-wide SNP arrays, roughly similar in average expected power to a conventional GWAS with 71% of the sample size (see Supplementary Note, Supplementary Figure 3, Supplementary Table 4 for results averaged by chromosome, minor allele frequency and coverage). We also quantified the genome-wide accuracy achieved by using all data from the whole exome scan (off-target and on-target); the average r2 increased to 0.77 when all data from the whole-exome study was used.
