Genotypes of DGN samples were genotyped on the Illumina HumanOmni 1-Quad BeadChip [15 (link)]. Nine hundred twenty-two samples have RNA-seq data available. We further removed related individuals and were left with 913 individuals. We imputed the genotypes on the Michigan Imputation Server [38 (link)]. We kept only SNPs with genotyping rate > 99%, minor allele frequency > 5%, and Hardy-Weinberg equilibrium < 10−6 using PLINK 2.0 [39 ].
RNA-seq reads were mapped to the reference genome (NCBI v37) by TopHat in ref. [15 (link)]. We further discarded reads that are mapped to multiple locations and reads with > 2 mismatches. Next, we removed genomic regions with low mappability. We quantified expression with HTseq [40 ]. Expression levels of 13,634 genes with at least 1CPM in at least 50% of the individuals were quantified. Finally, expression levels of these genes are quantified as Transcripts Per Million (TPM). We first quantile normalized the expression levels across samples. Then, we quantile normalized the expression levels to standard normal across genes before running testing for trans signals.
RNA-seq reads were mapped to the reference genome (NCBI v37) by TopHat in ref. [15 (link)]. We further discarded reads that are mapped to multiple locations and reads with > 2 mismatches. Next, we removed genomic regions with low mappability. We quantified expression with HTseq [40 ]. Expression levels of 13,634 genes with at least 1CPM in at least 50% of the individuals were quantified. Finally, expression levels of these genes are quantified as Transcripts Per Million (TPM). We first quantile normalized the expression levels across samples. Then, we quantile normalized the expression levels to standard normal across genes before running testing for trans signals.
Free full text: Click here
