Single Cell Clustering

What is single-cell clustering?

Single-cell clustering is a computational technique that groups individual cells into clusters based on their gene expression profiles, enabling the identification of cell types and states from scRNA-seq data.

Key Steps

  1. Data Preprocessing: Remove noise, correct artefacts, and normalise expression values.

  2. Dimensionality Reduction: Use PCA, t-SNE, or UMAP to reduce dimensionality.

  3. Clustering: Apply K-means, hierarchical, density-based (DBSCAN), or graph-based methods (Leiden) for grouping cells.

  4. Visualisation: Use UMAP plots to explore relationships between clusters.

Applications

  • Cell type identification — crucial for understanding tissue composition.

  • Disease subtyping — reveal disease-specific cell subpopulations for conditions such as cancer.

  • Developmental biology — track cell differentiation and developmental processes.

  • Immunology — identify immune cell populations, discover rare types.

  • Drug discovery — identify target cells and uncover resistance mechanisms.

  • Rare cell identification — detect and characterise rare populations.

scRNA-seq vs. Bulk RNA-seq

scRNA-seq vs Bulk RNA-seq

Image source: 10x Genomics

Property

scRNA-seq

Bulk RNA-seq

Resolution

Individual cells.

Average across all cells.

Heterogeneity

Detects cellular diversity.

Masks heterogeneity.

Applications

Complex tissues, rare cells.

Homogeneous samples.

Data complexity

High-dimensional, requires advanced methods.

Lower-dimensional, simpler to analyse.

Cost

Higher (single-cell isolation).

Lower (pooled RNA).

Sensitivity

Detects rare transcripts.

May miss rare transcripts.

In summary, scRNA-seq provides high-resolution insights into cellular heterogeneity, making it the foundation for methods like GARAGE that aim to model and augment this diversity.