Single-Cell Genomics: Approaches and Utility in Immunology

Abstract

Single-cell genomics offers powerful tools for studying immune cells, which make it possible to observe rare and intermediate cell states that cannot be resolved at the population level. Advances in computer science and single-cell sequencing technology have created a data-driven revolution in immunology. The challenge for immunologists is to harness computing and turn an avalanche of quantitative data into meaningful discovery of immunological principles, predictive models, and strategies for therapeutics. Here, we review the current literature on computational analysis of single-cell RNA-sequencing data and discuss underlying assumptions, methods, and applications in immunology, and highlight important directions for future research.

Keywords: dimensionality reduction; immune repertoire; single-cell RNA-sequencing; visualization.

Figure 1. (Key Figure): Single Cell RNA-Sequencing Analysis Outline

The standard workflow of single cell RNA-sequencing (scRNA-seq) studies begins with isolation of single cells; followed by library generation and high throughput RNA sequencing. Multiple computational and bioinformatics tools exist for quantifying gene expression. In general, these tools report the amount of reads that are associated with a specific gene, typically normalizing for gene length and library size. Before gene expression quantification, it is important to perform quality control analyses, including filtering for reads quality and eliminating cells with overall low library size. After quantification, batch effect, dropout effect and amplification bias should also be considered in normalizing scRNA-seq gene expression data (see Figure 2). Ultimately, scRNA-seq data can be subjected to dimensionality reduction algorithms to reveal different cell subpopulations, potentially infer their developmental trajectory during a dynamic process, and, if applicable, use sequencing data to identify their adaptive receptor sequence to assess questions of clonality and specificity. See Box 1 for an example of a typical processing pipeline.

Figure 2. Single Cell RNA-Sequencing Specific Challenges

Single cell RNA-seq (scRNA-seq) is by nature a fragile and sensitive technique. Experimental challenges inherent to this approach can confound data analysis and should be accounted for. First, technical batch effects can cause the data to artificially cluster by batch, overshadowing the biological variability. Second, dropout effect causes an enrichment of non-biological zero values that results in a bimodal expression profile, which can impair subpopulation analysis. Finally, certain genes may experience preferential gene amplification, introducing a bias that is difficult to adjust for. The use of UMI can help account for amplification bias and allow for more accurate quantification of starting RNA levels.