scCAD: Cluster decomposition-based anomaly detection for rare cell identification in single-cell expression data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for dissecting cellular heterogeneity in tissues.
• Identifying rare cell types is vital for understanding disease pathogenesis and biological processes.
• Current methods struggle to detect rare cells missed during initial clustering.

Purpose of the Study:

• To develop an advanced computational method for accurate rare cell type identification.
• To overcome limitations of existing clustering-based approaches in single-cell analysis.
• To enhance the discovery of critical cell populations in complex biological systems.

Main Methods:

• Propose Cluster decomposition-based Anomaly Detection (scCAD) for iterative cluster decomposition.
• Utilize differential gene expression signals within clusters to separate rare cell types.
• Benchmark scCAD against 10 state-of-the-art methods on 25 diverse scRNA-seq datasets.

Main Results:

• scCAD demonstrates superior performance in identifying rare cell types across multiple datasets.
• Case studies confirm scCAD's efficacy in complex scenarios like mouse and human tissues, and immunology data.
• The method successfully corrects rare cell type annotations and identifies disease-associated immune subtypes.

Conclusions:

• scCAD offers a robust and accurate solution for rare cell type identification in large-scale single-cell transcriptomics.
• This method provides valuable insights into disease progression by revealing critical cell populations.
• scCAD advances the application of scRNA-seq for biological discovery and clinical relevance.