XCVATR: Detection and Characterization of Variant Impact on the Embeddings of Single -Cell and Bulk RNA-Sequencing Samples

Abstract

Gene expression profiling via RNA-sequencing has become standard for measuring and analyzing the gene activity in bulk and at single cell level. Increasing sample sizes and cell counts provides substantial information about transcriptional architecture of samples. In addition to quantification of expression at cellular level, RNA-seq can be used for detecting of variants, including single nucleotide variants and small insertions/deletions and also large variants such as copy number variants. The joint analysis of variants with transcriptional state of cells or samples can provide insight about impact of mutations. To provide a comprehensive method to jointly analyze the genetic variants and cellular states, we introduce XCVATR, a method that can identify variants, detect local enrichment of expressed variants, within embedding of samples and cells. The embeddings provide information about cellular states among cells by defining a cell-cell distance metric. Unlike clustering algorithms, which depend on a cell-cell distance and use it to define clusters that explain cell clusters globally, XCVATR detects the local enrichment of expressed variants in the embedding space such that embedding can be computed using any type of measurement or method, for example by PCA or tSNE of the expression levels. XCVATR searches local patterns of association of each variant with the positions of cells in an embedding of the cells. XCVATR also visualizes the local clumps of small and large-scale variant calls in single cell and bulk RNA-sequencing datasets. We perform simulations and demonstrate that XCVATR can identify the enrichments of expressed variants. We also apply XCVATR on single cell and bulk RNA-seq datasets and demonstrate its utility.