ASURAT: functional annotation-driven unsupervised clustering of single-cell transcriptomes.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals cellular heterogeneity and dynamic transitions.
• Current gene-based analyses of scRNA-seq data demand extensive manual curation for biological interpretation.
• A need exists for efficient methods to annotate individual cells within complex datasets.

Purpose of the Study:

• To introduce ASURAT, a computational tool designed for simultaneous unsupervised clustering and functional annotation of scRNA-seq data.
• To enable the annotation of cell types, disease states, biological processes, and signaling pathways.
• To enhance the biological interpretability of complex and noisy transcriptomic data.

Main Methods:

• ASURAT employs correlation graph decomposition on genes within database-derived functional terms.
• The tool performs unsupervised clustering and functional annotation concurrently.
• It was validated using human peripheral blood mononuclear cell scRNA-seq datasets.

Main Results:

• ASURAT demonstrated improved usability and clustering performance compared to existing methods, requiring less manual curation.
• Application to human small cell lung cancer and pancreatic ductal adenocarcinoma datasets identified novel subpopulations and differentially expressed genes.
• The tool effectively dissects cell subpopulations and enhances biological interpretability.

Conclusions:

• ASURAT offers a powerful approach for analyzing single-cell transcriptomic data.
• It significantly reduces the manual effort required for interpreting scRNA-seq results.
• ASURAT facilitates deeper biological insights into cellular heterogeneity and function.