Estimating the Allele-Specific Expression of SNVs From 10× Genomics Single-Cell RNA-Sequencing Data

Abstract

With the recent advances in single-cell RNA-sequencing (scRNA-seq) technologies, the estimation of allele expression from single cells is becoming increasingly reliable. Allele expression is both quantitative and dynamic and is an essential component of the genomic interactome. Here, we systematically estimate the allele expression from heterozygous single nucleotide variant (SNV) loci using scRNA-seq data generated on the 10×Genomics Chromium platform. We analyzed 26,640 human adipose-derived mesenchymal stem cells (from three healthy donors), sequenced to an average of 150K sequencing reads per cell (more than 4 billion scRNA-seq reads in total). High-quality SNV calls assessed in our study contained approximately 15% exonic and >50% intronic loci. To analyze the allele expression, we estimated the expressed variant allele fraction (VAF_RNA) from SNV-aware alignments and analyzed its variance and distribution (mono- and bi-allelic) at different minimum sequencing read thresholds. Our analysis shows that when assessing positions covered by a minimum of three unique sequencing reads, over 50% of the heterozygous SNVs show bi-allelic expression, while at a threshold of 10 reads, nearly 90% of the SNVs are bi-allelic. In addition, our analysis demonstrates the feasibility of scVAF_RNA estimation from current scRNA-seq datasets and shows that the 3'-based library generation protocol of 10×Genomics scRNA-seq data can be informative in SNV-based studies, including analyses of transcriptional kinetics.

Keywords: RNA-seq; SNV; VAFRNA; genetic variation; monoallelic expression; sc-RNA-seq; sc-VAFRNA; single cell; single-cell RNA-sequencing.

Figure 1

Analytical workflow for an estimation of the variant allele fraction VAF_RNA from single-cell RNA-sequencing (sc-RNA-seq) data.

Figure 2

Number of genes, number of sequencing reads, and percent of mitochondrial genes for N8, N7, and N5 before (top) and after (bottom) the filtering out of cells with low-quality data.

Figure 3

Functional annotation (based on the position in regard to the harboring genes) of SNVs captured by the 10×Genomics platform with different required minimal counts of unique sequencing reads. At minR = 5, over 45% of the SNVs are positioned downstream of the 3′-UTR regions.

Figure 4

Histograms representing the distribution of scVAF_RNA at minR = 10 (top), minR = 5 (middle), and minR = 3 (bottom) for all the heterozygous SNVs in N8, N7, and N5. The bin width (x-axis) is 0.1; bin intervals are indicated in the middle of each plot. The y-axes show the numbers of VAF_RNA measurements in the individual cells. The total number of VAF_RNA estimations (n, across all the cells per group) is shown at the top of each histogram. The histograms are scaled in regard to the number of cells. Across the entire dataset, at minR = 10 and minR = 5, the majority of SNVs showed bi-allelic expression centered around a VAF_RNA value of 0.5 (0.4 < VAF_RNA < 0.6). In contrast, at minR = 3, the majority of SNVs presented with strict monoallelic expression (VAF_RNA = 0 or 1). The VAF_RNA distributions showed remarkable similarity across the three individuals (N8, N7, and N5).

Figure 5

scVAF_RNA estimated at positions covered by a minimum of 10 sequencing reads (top), 5 sequencing reads (middle), and 3 sequencing reads (bottom), across more than 1000 cells. For the majority of positions, VAF_RNA showed bi-allelic expression, with a substantial proportion of the scVAF_RNA estimations in the interval 0.4–0.6.

Figure 6

scVAF_RNA distribution at positions covered by a minimum of 10 sequencing reads (top), and three sequencing reads (bottom), across more than 1500 cells for genes reported by Borel et al. [9]. For the positions with minR = 10, no RME was suggested by the scVAF_RNA distribution for autosomal genes (i.e., for the majority of the estimations scVAF_RNA values were between 0.2 and 0.8), while positions covered with minR = 3 showed frequent monoallelic signals (scVAF_RNA > 0.8 or scVAF_RNA < 0.2). As expected, chrX shows strong RME patterns (see gene TSPAN6).

Figure 7

Percentage of cells (y-axis) displaying VAF_RNA < 0.2 (for each cluster of three, left), VAF_RNA between 0.2 and 0.8 (middle), and VAF_RNA > 0.8 (right); minR = 10. High concordance between the three donors is seen; SNVs on chromosome 1 are shown, and the results were similar genome-wide.

Figure 8

(a) Genes with multiple SNVs positioned in intronic and non-intronic sequences; high percentage of random monoallelic expression (RME) cells (VAF_RNA < 0.2 or > 0.8, y-axis) is obvious. (b) Average percentage of cells (y-axis) with bi-allelic expression across all SNVs in our dataset stratified by position in the gene; SNVs positioned in introns were be-allelic in a lower proportion of cells as compared to all other SNVs.