Single-Cell RNA-Seq: Understanding Cellular Heterogeneity

Traditional bulk RNA-Seq measures average gene expression across millions of cells. But tissues are heterogeneous — a tumor contains cancer cells, immune cells, fibroblasts, and endothelial cells, each with distinct transcriptional profiles. Single-cell RNA-Seq (scRNA-Seq) resolves this heterogeneity by profiling gene expression in individual cells.

How scRNA-Seq Works

The most popular platform is 10x Genomics Chromium, which uses microfluidic droplets to encapsulate individual cells with barcoded beads. Each cell receives a unique molecular barcode, allowing reads from the same cell to be grouped together computationally. A single run can profile 10,000 to 100,000 cells.

Other technologies include Smart-seq2, which provides full-length transcript coverage but lower throughput, and sci-RNA-seq3, which uses combinatorial indexing to profile millions of cells without droplet-based isolation.

The Analysis Pipeline

Cell Ranger from 10x Genomics handles the initial processing: demultiplexing, alignment, barcode counting, and generating the count matrix. This matrix is typically sparse — most genes in most cells have zero counts — due to the shallow sequencing depth per cell.

Scanpy (Python) and Seurat (R) are the two dominant analysis frameworks. The standard workflow involves quality filtering (removing doublets and dead cells), normalization, highly variable gene selection, dimensionality reduction (PCA), clustering (Leiden or Louvain algorithm), and visualization (UMAP or t-SNE).

Cell Type Annotation

After clustering, the key biological question is: what cell type does each cluster represent? This is done by examining marker genes — genes known to be specifically expressed in certain cell types. For example, CD3E marks T cells, CD19 marks B cells, and EPCAM marks epithelial cells.

Automated annotation tools like CellTypist and SingleR compare your clusters against reference atlases, providing probabilistic cell type assignments. The Human Cell Atlas project is building comprehensive reference datasets for every tissue in the human body.

Beyond Clustering

Advanced analyses include trajectory inference (pseudotime), which orders cells along developmental or differentiation paths; RNA velocity, which predicts the future state of cells based on the ratio of unspliced to spliced mRNA; and cell-cell communication analysis using tools like CellChat, which infers signaling interactions between cell types.