SMNN: batch effect correction for single-cell RNA-seq data via supervised mutual nearest neighbor detection.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Integrating single-cell RNA sequencing (scRNA-seq) data across multiple batches is crucial for comprehensive analysis.
• Existing batch effect correction methods often overlook valuable single-cell cluster label information.
• This oversight limits correction effectiveness, especially when biological variation overlaps with batch effects.

Purpose of the Study:

• To introduce SMNN (Supervised Mutual Nearest Neighbor), a novel method for scRNA-seq batch effect correction.
• To leverage single-cell cluster labels for improved data integration and biological accuracy.
• To demonstrate SMNN's superior performance compared to existing state-of-the-art methods.

Main Methods:

• Developed SMNN utilizing supervised mutual nearest neighbor detection.
• Incorporated single-cell cluster label information into the batch effect correction framework.
• Conducted extensive evaluations on both simulated and real-world scRNA-seq datasets.

Main Results:

• SMNN achieved superior merging of corresponding cell types across batches compared to MNN, Seurat v3, and LIGER.
• The method significantly reduced artificial differentiation between cell types originating from different batches.
• SMNN demonstrated enhanced retention of cell-type-specific features, improving the biological relevance of differentially expressed genes by up to 841.0%.

Conclusions:

• SMNN offers a more effective approach to batch effect correction in scRNA-seq data by integrating cluster label information.
• The method improves data integration quality, leading to more biologically meaningful downstream analyses.
• SMNN represents a significant advancement for researchers working with multi-batch scRNA-seq datasets.