Influences of rare protein-coding genetic variants on the human plasma proteome in 50,829 UK Biobank participants

Abstract

Combining human genomics with proteomics is becoming a powerful tool for drug discovery. Associations between genetic variants and protein levels can uncover disease mechanisms, clinical biomarkers, and candidate drug targets. To date, most population-level proteogenomic studies have focused on common alleles through genome-wide association studies (GWAS). Here, we studied the contribution of rare protein-coding variants to 1,472 plasma proteins abundances measured via the Olink Explore 1536 assay in 50,829 UK Biobank human exomes. Through a variant-level exome-wide association study (ExWAS), we identified 3,674 rare and significant protein quantitative trait loci (pQTLs), of which 76% were undetected in a prior GWAS performed on the same cohort, and we found that rare pQTLs are less likely to be random in their variant effect annotation. In gene-based collapsing analyses, we identified an additional 166 significant gene-protein pQTL signals that were undetected through single-variant analyses. Of the total 456 protein-truncating variant (PTV)-driven cis-pQTLs in the gene-based collapsing analysis, 99.3% were associated with decreased protein levels. We demonstrate how this resource can identify allelic series and propose biomarkers for several candidate therapeutic targets, including GRN, HSD17B13, NLRC4, and others. Finally, we introduce a new collapsing analysis framework that combines PTVs with missense cis-pQTLs that are associated with decreased protein abundance to bolster genetic discovery statistical power. Our results collectively highlight a considerable role for rare variation in plasma protein abundance and demonstrate the utility of plasma proteomics in gene discovery and unravelling mechanisms of action.

Genetic architecture of telomere length in 462,675 UK Biobank whole-genome sequences

“…14 ). To investigate this further, we performed somatic variant calling in 15 established CH and myeloid cancer driver genes ( Supplementary Table 12 ) using the complementary UK Biobank higher coverage exome sequencing data (Dhindsa et al 2022). Using these somatic CH calls, and adjusting for age, sex and smoking status, we performed collapsing analyses with our PC1 metric and replicated the previously described association between overall CH and shorter TL (Nakao et al 2022) ( Figure 3A ).…”

“…To detect putative clonal haematopoiesis, we used the pipeline described in Dhindsa et al(Dhindsa et al 2022). Briefly, using the same GRCh38 genome reference aligned reads as for WES germline variant calling, we ran somatic variant calling with GATK’s Mutect2 (v.4.2.2.0), After QC we focussed on a set of 15 genes ( Supplementary Table 12 ) exhibiting age dependent prevalence for further analyses including only PASS variant calls with 0.03 ≤ Variant Allele Frequency (VAF) ≤ 0.4 and Allelic Depth (AD) ≥ 3 across an annotated set of variants.…”

Multi-layered genetic approaches to identify approved drug targets

“…Third, we only focus on common genetic variants when associating transcript and protein levels. With the advent of coupled rare variant-protein level data, either from populations enriched for rare variants or sequencing data [14, 44], more powerful QTL-GWAS methods are likely to emerge that combine mechanistic insights gained from QTL approaches while capturing rare variant associations previously missed. Fourth, drug target data are sparse which limits the statistical power in benchmarking analyses.…”

Applying Machine Learning on UK Biobank biomarker data empowers case-control discovery yield

“…Clonal haematopoiesis (CH) is the presence of clonal somatic mutations in the blood of individuals without a haematologic malignancy 53 , and these somatic variants can be detected with germline variant callers 54 . To avoid conflating somatic with germline variants, and age confounded disease associations, we removed from further analysis fifteen genes we previously identified in the UK Biobank as carrying clonal somatic mutations 48 ( ASXL1, BRAF, DNMT3A, GNB1, IDH2, JAK2, KRAS, MPL, NRAS, PPM1D, PRPF8, SF3B1, SRSF2, TET2, TP53 ). A few other problematic genes as listed in Wang et al 2021 2 have also been removed from further investigation.…”