Mismapping of sequencing reads from polymorphic duplications generates spurious trans-eQTLs

The discovery of trans-acting expression quantitative trait loci (trans-eQTLs) remains plagued by false positives. One unaddressed source of these false positives is polymorphic duplications absent in the reference genome. Specifically, RNA sequencing (RNA-seq) reads from a non-reference gene duplicate have the potential to erroneously map to the single reference copy of the gene. These mismapped RNA-seq reads could generate spurious trans-eQTLs that localize near the insertion site of the non-reference duplicate but appear to regulate the reference copy of the gene. In such scenarios, similarly mismapped DNA-seq reads would create apparent long-range linkage disequilibrium between the spurious trans-eQTL and its target gene's locus. We call this footprint "mismapping linkage disequilibrium" and use it to flag spurious trans-eQTLs. We applied this strategy to published trans-eQTLs from Genotype-Tissue Expression (GTEx) and newly generated trans-eQTLs from MAGE, both already filtered for calls attributable to cross-mappability of RNA-seq reads between reference duplicates. This strategy demonstrated that at least 15.4% and 6.1% of trans-eQTLs in GTEx and MAGE, respectively, are spurious due to non-reference polymorphic duplications. We find that spurious trans-eQTLs are more likely to have large effect sizes. Thus, we expect large-sample studies such as eQTLGen (n > 30,000), which are powered to detect many more small-effect trans-eQTLs, to harbor smaller proportions of spurious trans-eQTLs than smaller studies such as GTEx (n = 838) and MAGE (n = 731). The polymorphic duplication of DUSP22 demonstrates how mismapping linkage disequilibrium can uncover both false-positive trans-eQTLs and trans-acting splicing QTLs (trans-sQTLs). Overall, our study offers a practical metric to improve trans-eQTL filtering.