We used the default versions of code to process our datasets. For genotype data, we applied PLINK v1.07 for QC to filter out those SNPs with genotype call rate <=95%, MAF <=0.01, misshap test <1×10−9, and a Hardy-Weinberg P>=0.001. Based on these genotyped information, we used the Beagle software v3.3.2 with the 1000 Genomes Project (2011 Phase 1b data freeze) and Minimac3 & Eagle v2.3 with the Haplotype Reference Consortium (HRC) reference panel of Caucasian ancestry v.1.1 to yield imputed dosage information of genotypes. For the whole genome sequencing project, we used BWA-mem v0.7.15 for the alignment and GATK v3.5 for the genotype calling. RNAseq dataset were aligned by Tophat v2.0 and v2.1 and transcript enrichments were estimated by RSEM package. The ChipSeq data were aligned by the BWA algorithm and peaks were detected by MACS2. Quality metrics of the above mentioned sequencing data were provided by Picard, which were also used to mark duplicated reads. Within-batch normalization was conducted through quantile normalization while the between-batches normalization was conducted through COMBAT.
