Rare variants regulate expression of nearby individual genes in multiple tissues

Abstract

The rapid decrease in sequencing cost has enabled genetic studies to discover rare variants associated with complex diseases and traits. Once this association is identified, the next step is to understand the genetic mechanism of rare variants on how the variants influence diseases. Similar to the hypothesis of common variants, rare variants may affect diseases by regulating gene expression, and recently, several studies have identified the effects of rare variants on gene expression using heritability and expression outlier analyses. However, identifying individual genes whose expression is regulated by rare variants has been challenging due to the relatively small sample size of expression quantitative trait loci studies and statistical approaches not optimized to detect the effects of rare variants. In this study, we analyze whole-genome sequencing and RNA-seq data of 681 European individuals collected for the Genotype-Tissue Expression (GTEx) project (v8) to identify individual genes in 49 human tissues whose expression is regulated by rare variants. To improve statistical power, we develop an approach based on a likelihood ratio test that combines effects of multiple rare variants in a nonlinear manner and has higher power than previous approaches. Using GTEx data, we identify many genes regulated by rare variants, and some of them are only regulated by rare variants and not by common variants. We also find that genes regulated by rare variants are enriched for expression outliers and disease-causing genes. These results suggest the regulatory effects of rare variants, which would be important in interpreting associations of rare variants with complex traits.

Fig 1. Power comparison between LRT-q and seven existing methods on simulated data.

A. for different effect sizes and fixed causal ratio (10%), and for fixed effect sizes (B. ≤ 0.99, C. ≤ 2.97, D. ≤ 4.95) and various causal ratios. Significance level α = 0.05.

Fig 2. RV eGenes detected from 49 tissues in the GTEx v8 dataset.

A. The relationship between the number of total RV eGenes detected by LRT-q in each tissue and the sample size of each tissue. The colors of the data points are randomly assigned. Each tissue has its own color. B. The number of total RV eGenes detected by each method. In panel B, only tissues with more than one RV eGene detected by any methods are included.

Fig 3. Tissue-sharing patterns of RV eGenes in the GTEx v8 dataset.

A. Pairwise tissue-sharing matrix of RV eGenes. It shows the fraction of shared RV eGenes in each pair of tissues. Here we use FDR < 10% to increase the number of RV eGenes. Tissues are sorted by clustering. Only tissues with more than one RV eGenes are included. B. The proportion of RV eGenes and CV eGenes shared among different numbers of tissues. Only tissues with more than 200 RV eGenes are selected. It shows the proportion of eGenes that are only detected in one tissue, in 2–4 tissues, and in more than 4 tissues.

Fig 4. Outlier analysis of RV eGenes detected by LRT-q in GTEx v8.

A. Enrichment of proximal rare variants in outliers compared to non-outliers for RV eGenes in each tissue. Tissues without RV eGenes are excluded. B. Enrichment of RV eGenes in disease-associated genes and genes related to common traits (BMI and Height) from public databases. The numbers represent p-values. In both panels, we show the mean values as dots and 95% confidence intervals as error bars.