A comparative study of cell type annotation methods for immune cells using single-cell sequencing technology

Abstract

Abstract: Single-cell sequencing is an emerging technique that allows high-throughput data analysis at an individual cell resolution and is applied in diverse fields of biology. Due to the large amount of data, downstream analysis is very complicated, and cell type annotation is a critical step; however, it is currently difficult to obtain good results. Single-cell sequencing has resulted in new breakthroughs in multi-omics, such as CITE-seq (Cellular Indexing of Transcriptomes and Epitopes by sequencing), which allows the measurement of surface marker proteins simultaneously with the sequencing of mRNA at the single-cell level. In this study, a CITE-seq dataset of human PBMCs (peripheral blood mononuclear cells) was annotated using the most popular reference-based annotation methods, including SingleR, Seurat with the RNA-seq dataset, and Seurat with both the RNA and protein datasets; the results were then compared with RNA and protein expression levels to determine the role of proteins in cell annotation. The results indicate that protein expression can supplement datasets with some mRNAs with low expression to improve accuracy. With the verification of single-cell biomarkers, the multi-omics annotation method Seurat with both the RNA and protein databases showed the best performance, especially in the differentiation of NK cells and T cells and of dendritic cells and monocytes. This study shows the significance of multi-omics information for improving cell annotation and has great potential for perfecting these annotations with more data support.