CDSKNN^XMBD: a novel clustering framework for large-scale single-cell data based on a stable graph structure.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Accurate cell grouping is crucial for single-cell RNA sequencing (scRNA-seq) data analysis.
• Existing clustering methods face challenges with large-scale and imbalanced cell type datasets.
• There is a need for efficient and accurate scRNA-seq clustering solutions.

Purpose of the Study:

• To introduce a novel single-cell clustering framework, CDSKNN^XMBD.
• To address limitations of current methods in handling large-scale and imbalanced scRNA-seq data.
• To achieve accurate and fast cell type identification.

Main Methods:

• Developed CDSKNN^XMBD, integrating partition and community detection algorithms.
• Utilized a stable K-nearest neighbor graph structure for clustering.
• Evaluated performance on diverse scRNA-seq datasets, including the human fetal atlas.

Main Results:

• CDSKNN^XMBD effectively clusters imbalanced single-cell data.
• Demonstrated high applicability and robustness across various datasets and sequencing techniques.
• Achieved significant efficiency gains, clustering 1.46 million cells in an average of 6.33 minutes, saving 33.3% to 99% of running time.

Conclusions:

• CDSKNN is a flexible and resilient clustering tool for scRNA-seq data.
• Particularly suitable for imbalanced datasets and large-scale analyses.
• Offers a promising solution for efficient and accurate cell type grouping.