A Single-Cell Sequencing Guide for Immunologists

Abstract

In recent years there has been a rapid increase in the use of single-cell sequencing (scRNA-seq) approaches in the field of immunology. With the wide range of technologies available, it is becoming harder for users to select the best scRNA-seq protocol/platform to address their biological questions of interest. Here, we compared the advantages and limitations of four commonly used scRNA-seq platforms in order to clarify their suitability for different experimental applications. We also address how the datasets generated by different scRNA-seq platforms can be integrated, and how to identify unknown populations of single cells using unbiased bioinformatics methods.

Keywords: 10X genomics chromium; MARS-seq; SMART-seq; dendritic cells; fluidigm C1; immunology; single-cell RNA sequencing.

Figure 1

Identification of cell types using scRNA-seq data from 10X Genomics Chromium system. (A) tSNE clustering of single cells in PBMC. (B) Alignment of clusters to known immune cell populations. (C) tSNE clustering of combined cluster 9 and 10 which was inferred as monocytes and DC. (D) Superimposed correlation-inferred cell type on the tSNE representation of combined cluster 9 and 10. (E) Superimposed CIBERSORT-based cell type classification on the tSNE representation of combined cluster 9 and 10.

Figure 2

Identification of cell types using scRNA-seq data from SMART-seq2. (A) tSNE clustering of dendritic cell subsets. (B) Superimposed CIBERSORT-based cell type classification on the tSNE representation of SMART-seq2 dataset. (C) Alignment of SMART-seq2 clusters with microarray dataset of DC subsets. (D) tSNE clustering of DC cluster derived from 10X Genomics Chromium dataset. (E) Superimposed CIBERSORT-based cell type classification on the tSNE representation of DC cluster derived from 10X Genomics Chromium dataset. (F) Alignment of DC clusters with microarray dataset of DC subsets.

Figure 3

Batch effect correction of SMART-seq2 dataset. (A) Batch effect was observed in two separate SMART-seq2 datasets before CCA normalization, but this was absent after application of CCA normalization. (B) Cell clusters corresponded to the batch of SMART-seq2 dataset before CCA normalization. After CCA normalization was applied, both batches of single cells overlapped with each other.

Figure 4

Correction of technical variation in DC subset dataset from 10X Genomics Chromium and SMART-seq2 datasets. (A) tSNE clustering of SMART-seq2 and 10X Genomics Chromium dataset. (B) Cell type identification in the combined tSNE clusters of SMART-seq2 and 10X Genomics Chromium dataset. (C) CCA normalization of DC subsets from SMART-seq2 and 10X Genomics Chromium dataset. (D) Identification of cell types after CCA normalization.

Figure 5

Correction of technical variation in monocytes and DC subset dataset from 10X Genomics Chromium and SMART-seq2 datasets. (A) tSNE clustering of SMART-seq2 and 10X Genomics Chromium datasets. (B) Cell type identification in the combined tSNE clusters of SMART-seq2 and 10X Genomics Chromium datasets. (C) CCA normalization of monocytes and DC subsets from SMART-seq2 and 10X Genomics Chromium datasets. (D) Identification of cell types after CCA normalization.