Sequencing and read mapping

Libraries were sequenced on the Illumina HiSeq 2000 in 2 x 101 bp mode following the standard TruSeq SBS protocol. The eight HapMap samples were sequenced on a single lane, with a mean of 18.3 M paired end reads per sample. In the Argentine study, the two pooled libraries were sequenced on four and five separate lanes respectively, with a mean of 17.5 M reads per sample. Reads were mapped to the human reference genome (build 37) with BWA[23] with the -q 20 parameter to include soft clipping of low quality bases. Local realignment of reads around known indels and base quality recalibration were performed with GATK[18]. We defined the target region for the HapMap samples by taking the union of predicted restriction site fragments between 400–700 bp that had ≥ 3X mean coverage, and where ≥ 10% of reads had a mate pair mapped to a restriction site (S2A Fig). The target region for the Argentine samples was defined in the same way, but with predicted fragments between 200–600 bp. Argentine samples with < 30% of reads mapped to restriction sites (26/89 samples) were excluded from further analyses.