A cell marker-based clustering strategy (cmCluster) for precise cell type identification of scRNA-seq data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell transcriptome analysis is crucial for understanding cell function diversity.
• Cell clustering and population recognition are key challenges in this field.
• Current methods often analyze clustering precision and biological explanation separately.

Purpose of the Study:

• To develop a flexible strategy balancing clustering accuracy and biological explanation for single-cell data.
• To enhance cell population recognition and functional interpretation.
• To provide a comprehensive approach for analyzing single-cell transcriptome data.

Main Methods:

• Introduced cmCluster, a modified Louvain clustering method.
• Utilized genetic algorithm (GA) and grid search for optimal cluster identification.
• Integrated cell type annotation results into the clustering process.

Main Results:

• cmCluster effectively recognized cell populations and explained biological functions, even with incomplete data or multiple sources.
• The method generated distinct cell clusters with clear boundaries and identified potential marker genes.
• Demonstrated utility in identifying appropriate subtypes within complex datasets.

Conclusions:

• cmCluster offers researchers effective screening strategies for improved biological analysis.
• The approach helps reduce artificial bias in single-cell data interpretation.
• Facilitates comparison and analysis across multiple single-cell studies.