CIDER: an interpretable meta-clustering framework for single-cell RNA-seq data integration and evaluation.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Clustering joint single-cell RNA sequencing (scRNA-Seq) data faces challenges from batch effects and biological variability.
• Current methods for batch effect removal often rely on strict assumptions about cell population similarity across samples.

Purpose of the Study:

• To introduce CIDER, a meta-clustering workflow designed to overcome limitations in scRNA-Seq data integration.
• To demonstrate CIDER's superior performance compared to existing scRNA-Seq clustering and integration methods.
• To showcase CIDER's utility in evaluating the biological validity of data integration without prior cell annotations.

Main Methods:

• Developed CIDER, a meta-clustering workflow utilizing inter-group similarity measures.
• Evaluated CIDER on both simulated and real-world scRNA-Seq datasets.
• Compared CIDER's performance against established scRNA-Seq clustering and integration approaches.

Main Results:

• CIDER demonstrated superior performance in clustering joint scRNA-Seq data compared to other methods.
• The workflow effectively handled confounding factors like batch effects and biological variability.
• CIDER successfully assessed the biological correctness of data integration in real datasets.

Conclusions:

• CIDER offers a robust and flexible approach for clustering and integrating scRNA-Seq data.
• The method enhances the reliability of scRNA-Seq data analysis by mitigating batch effects.
• CIDER provides a valuable tool for evaluating data integration quality without requiring pre-existing cell type information.