Improving Single-Cell RNA-seq Clustering by Integrating Pathways.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables high-resolution analysis of cellular heterogeneity.
• Clustering algorithms are crucial for identifying cell populations in scRNA-seq data.
• Noise and data sparsity can compromise the accuracy and robustness of existing clustering methods.

Purpose of the Study:

• To investigate the impact of integrating biological pathway information on the performance of scRNA-seq clustering.
• To evaluate whether pathway integration can enhance the accuracy and robustness of state-of-the-art clustering methods.

Main Methods:

• Evaluated 10 single-cell clustering methods across 26 scRNA-seq datasets.
• Integrated pathway information using the AUCell method and similarity network fusion.
• Assessed performance using three accuracy metrics, three noise generation strategies, and robustness indicators.

Main Results:

• Pathway integration significantly improved the accuracy of most tested single-cell clustering methods.
• Robustness against noise was substantially enhanced by incorporating pathway data.
• The proposed framework demonstrated consistent performance improvements across diverse datasets.

Conclusions:

• Integrating pathway information is a powerful strategy to mitigate noise and boost performance in scRNA-seq data analysis.
• This approach offers a reliable way to improve the biological interpretability of single-cell clustering results.
• The findings suggest that pathway-aware clustering should be a standard consideration in scRNA-seq data analysis.