A Strategy to Compare Single-Cell RNA Sequencing Data Sets Provides Phenotypic Insight into Cellular Heterogeneity...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) generates high-dimensional transcriptomic data.
• Comparing scRNA-seq datasets is challenging due to cellular heterogeneity.
• Accurate comparison is crucial for understanding biological similarities and differences between samples.

Purpose of the Study:

• To introduce scCompare, a novel computational pipeline for comparing scRNA-seq datasets.
• To enable robust evaluation of biological similarities and differences in high-dimensional single-cell data.
• To facilitate novel cell type detection by allowing unmapped cells.

Main Methods:

• scCompare uses correlation-based mapping to transfer phenotypic identities between datasets.
• It averages transcriptomic signatures from annotated cell clusters.
• Statistically derived cutoffs allow for unmapped cells, aiding novel cell type discovery.

Main Results:

• scCompare demonstrated higher precision and sensitivity than single-cell variational inference (scVI) on human peripheral blood mononuclear cells (PBMCs).
• It confirmed a distinct cell cluster in a cardiomyocyte differentiation dataset, differing between protocols.
• Analysis of cell atlas data revealed insights into cellular heterogeneity and biological diversity.

Conclusions:

• scCompare is a valuable tool for comparing large scRNA-seq datasets.
• The pipeline accurately identifies cell types and outperforms existing methods.
• scCompare facilitates the discovery of novel cell types and enhances understanding of biological diversity.