Human genomics. Effect of predicted protein-truncating genetic variants on the human transcriptome

Abstract

Accurate prediction of the functional effect of genetic variation is critical for clinical genome interpretation. We systematically characterized the transcriptome effects of protein-truncating variants, a class of variants expected to have profound effects on gene function, using data from the Genotype-Tissue Expression (GTEx) and Geuvadis projects. We quantitated tissue-specific and positional effects on nonsense-mediated transcript decay and present an improved predictive model for this decay. We directly measured the effect of variants both proximal and distal to splice junctions. Furthermore, we found that robustness to heterozygous gene inactivation is not due to dosage compensation. Our results illustrate the value of transcriptome data in the functional interpretation of genetic variants.

Fig. 1

Schematic overview of the study. We prepared an integrated DNA and RNA sequencing data set by combining the pilot phase of the GTEx project of 173 individuals with up to 30 tissues per individual (total = 1634 samples) and the Geuvadis project of lymphoblastoid cell line (LCL) DNA and RNA sequencing in 462 individuals. From these data, we analyzed the effect of predicted proteintruncating genetic variants on the human transcriptome, including: a) nonsense single nucleotide variants (SNVs); b) frameshift indels; c) large deletion variants; and d) splice-disrupting SNVs.

Fig. 2

Allele-specific expression analysis. A. Proportion of rare SNVs with allele-specific expression (ASE) for synonymous variants (n = 25,233) and nonsense variants predicted to escape (n = 158) or trigger (n = 287) nonsense-mediated decay (NMD). B. Proportion of rare indels with ASE for inframe (n = 355) and frameshift indel variants predicted to escape (n = 77) or trigger (n = 129) NMD. Due to different quality filters, the proportions are not directly comparable to those in panel A. C. ROC curve for predicting NMD with binary classification defined as no ASE (= escape) and moderate, strong, or heterogeneous ASE (= trigger). The filled circles show the specificity and sensitivity for NMD prediction with alternative simple distance rules (inset). D. Multi-tissue ASE classification for rare nonsense variants predicted to trigger NMD (n = 287). E. Example of ASE data across 6 tissues for a heterozygous carrier of the nonsense variant rs149244943 in gene PHKB (phosphorylase kinase, beta) classified as having heterogeneous ASE effects across the seven tissues. We confirmed that this effect is not driven by a common tissue-specific eQTL. F. Example of ASE data across 16 tissues for a heterozygous carrier of the nonsense variant rs119455955, a disease mutation for recessive late-infantile neuronal ceroid lipofuscinosis in gene TPP1 (tripeptidyl peptidase I), classified as having moderate ASE across all tissues. For all plots 95% confidence intervals are shown.

Fig. 3

Splicing disruption. A. Proportion of variants disrupting splicing at each distance +/− 25bp from donor and acceptor site, (* P < 0.05, ** P < 0.01, *** P < 0.001; green for P < 0.05; upper limit of 95% CI is shown; P value evaluated using the estimated proportion of variants supporting the alternative distribution x the effect size of the alternative distribution). B. Classification of splice disruption events: exon skipping (low exon quantification value, no impact on intron quantification), exon elongation (high intron quantification value, no impact on exon quantification), and mixture (high intron and low exon quantification values). C. Diagram of donor and acceptor splice junctions and sequence logo of represented sequences. D. Effect size estimates (in standard deviations from the population distribution; 95% CI is shown) of the variants on splice junction quantification value. E. Median GERP of all variants F. Distribution of common variants identified in an independent exome sequencing study of 4,500 Swedish individuals. G. Distribution of reported disease-causing variants in ClinVar.