Single-cell RNA-seq data clustering: A survey with performance comparison study.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Clustering analysis is crucial for identifying cell types and states in single-cell RNA sequencing (scRNA-seq) data.
• Existing algorithms vary in techniques, evaluation datasets, and performance metrics, hindering objective method selection.
• A clear understanding of algorithm strengths and weaknesses is needed for effective scRNA-seq data analysis.

Purpose of the Study:

• To provide a comprehensive review of major scRNA-seq data clustering methods.
• To conduct a rigorous performance comparison of state-of-the-art scRNA-seq clustering algorithms.
• To guide researchers in selecting appropriate algorithms for their specific scRNA-seq datasets.

Main Methods:

• Reviewed 13 state-of-the-art scRNA-seq data clustering algorithms.
• Collected and utilized 12 publicly available real scRNA-seq datasets for evaluation.
• Compared algorithm performance from multiple perspectives, considering diverse data structures.

Main Results:

• Significant performance diversity was observed among the evaluated scRNA-seq clustering algorithms.
• Even top-performing algorithms showed limitations, particularly on datasets with complex cellular structures.
• No single algorithm consistently excelled across all tested datasets.

Conclusions:

• Current scRNA-seq clustering methods exhibit considerable variability in performance.
• Further research is essential to develop more robust, accurate, and efficient clustering algorithms for scRNA-seq data.
• The findings highlight the need for algorithm development tailored to complex biological data structures.