Having used 'bcftools call' to extract sites and alleles, we had a set of baseline non-LD genotype calls (see Genotype likelihood calculations section). Based on these calls for chromosome 22, some outlier samples were evident and we removed 150 samples showing evidence for fewer than 10,000 non-reference SNPs or more than 10 singletons across the chromosome. This left a total of 32,920 samples.