IDclust: Iterative clustering for unsupervised identification of cell types with single cell transcriptomics and...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell data analysis requires robust cell type characterization.
• Current clustering methods lack biological relevance and hierarchical structure.
• Manual parameter optimization hinders reproducibility and scalability.

Purpose of the Study:

• To develop a novel unsupervised clustering framework, IDclust, for systematic cell type identification.
• To enable the discovery of biologically meaningful clusters and hierarchical relationships.
• To improve the accuracy and interpretability of single-cell data analysis.

Main Methods:

• IDclust utilizes biologically relevant thresholds (fold change, adjusted P-value, fraction of expressing cells).
• The framework iteratively clusters data subsets, ensuring significant feature differences.
• Hierarchical relationships are established by processing clusters at multiple resolutions.

Main Results:

• IDclust demonstrates superior clustering accuracy on reference single-cell transcriptomic datasets.
• The method successfully identified previously unannotated cell populations.
• Branching patterns in scATAC-seq data and multi-omic relationships were elucidated.

Conclusions:

• IDclust automates single-cell data exploration and cell type annotation.
• The framework provides a biologically interpretable basis for clustering.
• IDclust offers a versatile tool for multi-omic single-cell analyses.