Transcriptomic signatures across human tissues identify functional rare genetic variation

Abstract

Rare genetic variants are abundant across the human genome, and identifying their function and phenotypic impact is a major challenge. Measuring aberrant gene expression has aided in identifying functional, large-effect rare variants (RVs). Here, we expanded detection of genetically driven transcriptome abnormalities by analyzing gene expression, allele-specific expression, and alternative splicing from multitissue RNA-sequencing data, and demonstrate that each signal informs unique classes of RVs. We developed Watershed, a probabilistic model that integrates multiple genomic and transcriptomic signals to predict variant function, validated these predictions in additional cohorts and through experimental assays, and used them to assess RVs in the UK Biobank, the Million Veterans Program, and the Jackson Heart Study. Our results link thousands of RVs to diverse molecular effects and provide evidence to associate RVs affecting the transcriptome with human traits.

Detection of aberrant splicing events in RNA-seq data using FRASER

“…After quality filtering, this dataset consisted of 7,842 RNA-seq samples from 48 tissues of 543 assumed healthy donors. Although the GTEx donors did not suffer from any rare disease, the samples may present aberrant splicing events, just as they present genes with aberrant expression levels 21 , 27 . After filtering for expressed junctions per tissue (“Methods” section), the FRASER splice site map contained on average 137,058 (±5,848 standard deviation across tissues) donor sites and 136,743 (±5,920) acceptor sites (Supplementary Fig.…”

“…Based on the latent space, FRASER models the expected value of each observation (“Methods” section). In contrast to methods such as LeafCutter 22 , LeafCutterMD 20 , and SPOT 21 , we modeled each junction individually and did not model jointly all junctions of a gene. We considered the observations that significantly deviated from their expected value as outliers.…”

“…FRASER considers each junction of a gene individually. In principle, this can be less sensitive than considering all junctions of a gene in a joint model, as with the Dirichlet-Multinomial based methods LeafCutterMD 20 , SPOT 21 , and the LeafCutter adaptation of Kremer et al 17 . To assess this, we performed a benchmark whereby splicing outliers are simulated by swapping out, for a single individual, skin and brain tissue read counts for a differentially alternatively spliced gene (“Methods” section).…”

“…Three distinct methods developed by (1) Cummings et al 16 , (2) Kremer et al 17 , and (3) Frésard et al 18 have been employed to call aberrant splicing in RNA-seq data for rare disease diagnostics. Moreover, two additional methods to call aberrant splicing were developed in parallel to this study: LeafCutterMD 20 and SPOT 21 . All five methods make use of the so-called RNA-seq split reads, whose ends align to two separated genomic locations of the same chromosome strand and are, therefore, evidence of splicing events.…”

“…Instead, Kremer et al 17 tested the significance of differential splicing using LeafCutter 22 , a multivariate count fraction model developed for mapping splicing quantitative trait loci. This approach, along with the more recent LeafCutterMD 20 and SPOT 21 , which are also multivariate approaches, allowed controlling for false discovery rate (FDR) and are less dependent on cohort size. One limitation, however, was made evident by Frésard et al 18 , who showed that strong covariations of split-read-based splicing metrics are widespread within RNA-seq compendia.…”