Improving expression estimates of HLA alleles in NGS data using personalized genomic sequences and its effect on eQTL mapping

Abstract

Abstract HLA genes are the most polymorphic in the human genome. Specific genetic variants in the HLA locus are frequently associated with diseases and other traits in GWAS studies. How much of those associations are due to functional differences in the peptide binding specificity of the various HLA genes encoded in the locus, vs. how many associations are due to differences in the expression level of the HLA genes has not been systematically determined. One hurdle in assessing HLA gene expression from RNA-Seq data when using standard analysis methods is the use of a single reference genome. We compared the estimated expression of HLA genes from 15 immune cell types of 91 individuals from the DICE database, using either the reference human genome sequence, or a personalized genome sequences based on HLA typing of the donors. This analysis showed that quantifying HLA gene expression using personalized genomic sequences allows to separate SNPs in association with HLA expression that are artifacts based on differences in read-mappability from other SNPs that are truly in association with personalized expression levels.