scBCN: deep learning-based batch correction network for integration of heterogeneous single-cell data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell data analysis faces challenges in correcting batch effects and identifying cell types accurately.
• Existing methods struggle to differentiate technical noise from true biological variation in heterogeneous datasets.

Purpose of the Study:

• To introduce single-cell Batch Correction Network (scBCN), a novel framework for robust batch effect correction.
• To preserve biological variability during data integration for improved cell type identification.

Main Methods:

• scBCN integrates inter-batch similar cluster identification with a deep residual neural network.
• Benchmarking experiments were conducted on simulated and real-world single-cell datasets.

Main Results:

• scBCN demonstrated superior performance in both batch correction and biological variation conservation compared to existing methods.
• The framework successfully integrated cross-species and cross-omics data.

Conclusions:

• scBCN offers a powerful solution for accurate cell type identification and analysis of complex single-cell datasets.
• The method has significant potential for advancing biomedical research through enhanced data integration and discovery.