Sequencing reads were aligned to the human reference using BWA-mem (version 0.7.15)19. Resulting BAM files contain all reads (passing or failing vendor quality checks), whether or not they aligned. Duplicate reads were detected and marked using Picard’s MarkDuplicates module (version 2.4.1) (http://broadinstitute.github.io/picard/). Local alignment was performed around indels to identify putative insertions or deletions in the region using the GATK20,21 (version 3.5) indel realignment tool. Base quality score recalibration was performed using the GATK BQSR. This step uses observed data to improve the quality scores for each base in the sequence. GATK HaplotypeCaller and GenotypeGVCFs modules were used to generate individual genotype calls in genomic VCF and VCF format. Following variant calling, we ran the variant quality recalibration step in the GATK pipeline to empirically calibrate high quality variants. To ensure a high level of accuracy in genotype calls from sequencing, variants were filtered for minimum read depth (DP), variant calling confidence score (QD), VQSLOD and mapping and variant quality scores (MQ, GQ). Variant-level QC was performed using PLINK15 which includes checking genotype concordance using previous GWAS data, excluding variants
