ascend: R package for analysis of single-cell RNA-seq data

Abstract

Background: Recent developments in single-cell RNA sequencing (scRNA-seq) platforms have vastly increased the number of cells typically assayed in an experiment. Analysis of scRNA-seq data is multidisciplinary in nature, requiring careful consideration of the application of statistical methods with respect to the underlying biology. Few analysis packages exist that are at once robust, are computationally fast, and allow flexible integration with other bioinformatics tools and methods.

Findings: ascend is an R package comprising tools designed to simplify and streamline the preliminary analysis of scRNA-seq data, while addressing the statistical challenges of scRNA-seq analysis and enabling flexible integration with genomics packages and native R functions, including fast parallel computation and efficient memory management. The package incorporates both novel and established methods to provide a framework to perform cell and gene filtering, quality control, normalization, dimension reduction, clustering, differential expression, and a wide range of visualization functions.

Conclusions: ascend is designed to work with scRNA-seq data generated by any high-throughput platform and includes functions to convert data objects between software packages. The ascend workflow is simple and interactive, as well as suitable for implementation by a broad range of users, including those with little programming experience.

Keywords: R package; clustering; data visualization; differential expression; filtering; normalization; scRNA-seq; single cell.

Figure 1

A summary of the typical analysis workflows and major function groups available in ascend.

Figure 2

Graphics generated by ascend during different stages of analysis. (A) Quality control plots include a boxplot representing distribution of library sizes across each batch, a boxplot representing the expression of the top 25 most abundant transcripts, and violin plots representing proportion of mitochondrial-related transcripts to total expression per sample. (B) Normalization quality control plot represents the expression of the SOX2 gene before and after RLE normalization. (C) Scree plot related to principal component dimensionality reduction. (D) Clustering plots include a cluster-labeled dendrogram and a line plot depicting the relationships between cluster numbers and stability.