These are subsets of the 1000G phase1 VCFs. For simplicity, the number of samples
has been reduced using bcftools, e.g.

$ bcftools view -sHG00531,HG00533,HG00534,HG00536,HG00537,HG00542,HG00543,HG00553,HG00554,HG00556,HG00557 $DATADIR/ALL.chr3.phase1_release_v3.20101123.snps_indels_svs.genotypes.vcf.gz | head -n 133 | grep -v "##contig" | bgzip > chr3.vcf.gz

We needed to remove the ##contig header because pyvcf didn't like it and said it was 
malformed.
