REBET: a method to determine the number of cell clusters based on batch effect removal

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Accurate cell cluster number determination is crucial for single-cell RNA sequencing (scRNA-seq) data analysis.
• Current methods struggle with expression variability and batch effects, limiting performance.
• Batch effects introduce unwanted variability from different experimental conditions or labs.

Purpose of the Study:

• To develop a novel method for accurately determining the number of cell clusters in scRNA-seq data.
• To address limitations in capturing expression variability and mitigating batch effects.
• To improve the robustness and accuracy of cell cluster identification.

Main Methods:

• Proposed REBET (Removal of Batch Effect and Testing) method.
• Partition cells into k clusters and subsequently remove batch effects among them.
• Evaluate batch effect removal quality using Average Range of Normalized Mutual Information (ARNMI).

Main Results:

• REBET accurately and robustly estimates the optimal number of cell clusters.
• The method demonstrated superior performance compared to state-of-the-art techniques.
• Validation performed on 32 simulated and 14 published scRNA-seq datasets.

Conclusions:

• REBET offers a significant advancement in determining cell cluster numbers for scRNA-seq data.
• The approach effectively handles batch effects, leading to more reliable results.
• This method enhances the accuracy and robustness of scRNA-seq data analysis.