Transcriptome-wide Mendelian Randomisation Using Expression, Protein and Splicing Quantitative Trait Loci to Identify Novel Drug Targets Associated With Risk of Gliomagenesis

Abstract

Abstract AIMS To use two-sample Mendelian Randomisation to prioritise novel genes for drug targeting in glioma risk using expression, protein and splicing quantitative trait loci (eQTLs, pQTLs and sQTLs, respectively). METHOD We used genetic variants from GWAS to compare the effects of eQTLs, pQTLs and sQTLs on genetic liability to glioma risk (n=12,496). eQTL data was retrieved from the MetaBrain study of five different brain tissues (n=108 to 2,970). pQTL data was retrieved from the BrainQTL study of dorsolateral prefrontal cortex tissue (n = 330), and sQTL data was retrieved from Genotype-Tissue Expression (GTEx) Project (v8) of 13 brain tissues (n=114 to 209). SNPs which were trans-QTLs (>1 Mb) due to risk of horizontal pleiotropy, and greater than GWAS significance (P > 5 × 10-8) were excluded. Instruments were constructed using independent SNPs (r2 < 0.001). Results underwent relevant sensitivity analysis; colocalisation and Steiger filtering. RESULTS We found 123 MR associations for 28 genes with genetically predicted altered expression, protein or splicing levels affected risk of glioma. These included genes recurrently associated with glioma risk such as TERT, RTEL1, CDKN2A/B and EGFR, and novel putative genes; BTN3A2, GALNT6, GMPPB and FAIM. Following sensitivity analyses, we found that 58 MR associations for 25 genes had robust evidence. CONCLUSION We identified four novel associations not previously found in GWAS; BTN3A2 (6p22.2), FAIM (3q22.3), GALNT6 (12q13.13) and GMPPB (3p21.31) are implicated in glioma risk and are also not located on a known glioma risk locus. Further analysis is required to identify the suitability of these genes for drug targeting in glioma.