Multi-level cellular and functional annotation of single-cell transcriptomes

Abstract

Single-cell RNA-sequencing (scRNA-seq) offers unprecedented insight into heterogenous biology, allowing for the interrogation of cellular populations and gene expression programs at single-cell resolution. Here, we introduce scPipeline, a single-cell analytic toolbox that offers modular workflows for multi-level cellular annotation and user-friendly analysis reports. Novel methods that are introduced to facilitate scRNA-seq annotation include: (i) co-dependency index (CDI)-based differential expression; (ii) cluster resolution optimization using a marker-specificity criterion; (iii) marker-based cell-type annotation with Miko scoring; and (iv) gene program discovery using scale-free shared nearest neighbor network (SSN) analysis. Our unsupervised and supervised procedures were validated using a diverse collection of scRNA-seq datasets and we provide illustrative examples of cellular and transcriptomic annotation of developmental and immunological scRNA-seq atlases. Overall, scPipeline provides a flexible computational framework for in-depth scRNA-seq analysis.