RNA sequencing variants are enriched for eQTL in cattle tissues

Abstract

AbstractAssociation testing between molecular phenotypes and genomic variants can help to understand how genotype affects phenotype. RNA sequencing provides access to molecular phenotypes such as gene expression and alternative splicing while DNA sequencing or microarray genotyping are the prevailing options to obtain genomic variants. Here we genotype variants for 74 male Braunvieh cattle from both DNA and deep total RNA sequencing from three tissues. We show that RNA sequencing calls approximately 40% of variants (7-10 million) called from DNA sequencing, with over 80% precision, rising to over 92% of variants called with nearly 98% precision in highly expressed coding regions. Allele-specific expression and putative post-transcriptional modifications negatively impact variant genotyping accuracy from RNA sequencing and contribute to RNA-DNA differences. Variants called from RNA sequencing detect roughly 75% of eGenes identified using variants called from DNA sequencing, demonstrating a nearly 2-fold enrichment of eQTL variants. We observe a moderate-to-strong correlation in nominal association p-values (Spearman ρ2∼0.6), although only 9% of eGenes have the same top associated variant. We also find several highly significant RNA variant-only eQTL, demonstrating that caution must be exercised beyond filtering for variant quality or imputation accuracy when analysing or imputing variants called from RNA sequencing.