Improved SNV discovery in barcode-stratified scRNA-seq alignments

Abstract

Single cell SNV analysis is an emerging and promising strategy to connect cell-level genetic variation to cell phenotypes. At the present, SNV detection from 10x Genomics scRNA-seq data is typically performed on the pooled sequencing reads across all cells in a sample. Here, we assess the gain of information of SNV assessments from individual cell scRNA-seq data, where the alignments are split by barcode prior to the variant call. For our analyses we use publicly available sequencing data on the human breast cancer cell line MCF7 cell line generated at consequent time-points during anti-cancer treatment. We analyzed SNV calls by three popular variant callers – GATK, Strelka2 and Mu-tect2, in combination with a method for cell-level tabulation of the sequencing read counts bearing SNV alleles – SCReadCounts. Our analysis shows that variant calls on individual cell alignments identify at least two-fold higher number of SNVs as compared to the pooled scRNA-seq. We demonstrate that scSNVs exclusively called in the single cell alignments (scSNVs) are substantially enriched in novel genetic variants and in coding functional annotations, in particular, stop-codon and missense substitutions. Furthermore, we find that the expression of some scSNVs correlates with the expression of their harbouring gene (cis-scReQTLs).Overall, our study indicates an immense potential of SNV calls from individual cell scRNA-seq data and emphasizes on the need of cell-level variant detection approaches and tools. Given the growing accumulation of scRNA-seq datasets, cell-level variant assessments are likely to significantly contribute to the understanding of the cellular heterogeneity and the relationship between genetics variants and functional phenotypes. In addition, cell-level variant assessments from scRNA-seq can be highly informative in cancer where they can help elucidate somatic mutations evolution and functionality.

A wealth of novel cell-specific expressed SNVs from tumor and normal scRNA-seq datasets

“…Variant calling from individual scRNA-seq alignments is a new and unexplored approach; therefore, to minimize false positives among the novel sceSNVs, we performed stringent quality filtering and examination of the sceSNV confidence at several levels. First, we used for our analyses the intersection of the highest quality calls in at least two cells per dataset by two callers widely used for RNA variant detection, GATK and Strelka2 (S_Methods) [18]. In parallel, for all novel sceSNV positions we estimated the variant read counts across all cells in each dataset using a method for cell-level tabulation of the sequencing read counts bearing reference and variant alleles from barcoded scRNA-seq alignments.…”

SCExecute: cell barcode-stratified analyses of scRNA-seq data

“…Methods for cell-specific scRNA-seq analysis have been focused on gene expression, where popular approaches – such as STARsolo and CellRanger – integrate alignment with concurrent barcode demultiplexing and the assignment of read counts to genes (Kaminow et al ., 2021; Tran et al ., 2019). Additional cell-level transcriptome feature analyses – for example, expressed genetic variation, allele-specific expression and splicing – are now beginning to emerge, creating new knowledge, and demonstrating the substantial information content of scRNA-seq data (Liu et al ., 2021; Prashant et al ., 2020; N. M. Prashant et al ., 2021; La Manno et al ., 2018).…”

SCExecute: custom cell barcode-stratified analyses of scRNA-seq data

“…Methods for cell-specific scRNA-seq analysis have been focused on gene expression, where popular approaches—such as STARsolo and CellRanger—integrate alignment with concurrent barcode demultiplexing and the assignment of read counts to genes ( Kaminow et al , 2021 ; Tran et al , 2019 ). Additional cell-level transcriptome feature analyses—for example, expressed genetic variation, allele-specific expression and splicing—are now beginning to emerge, demonstrating the substantial information content of scRNA-seq data ( La Manno et al , 2018 ; Larsson et al , 2021 ; Liu et al , 2021 ; Prashant et al , 2020 , 2021a , b ; Schnepp et al , 2019 ; Vu et al , 2019 ). These types of analyses can benefit from widely applicable cell-level read-stratifying methods.…”