NDMNN: A novel deep residual network based MNN method to remove batch effects from scRNA-seq data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) generates large datasets vital for biological research.
• Batch effects, arising from experimental variations, can confound scRNA-seq data analysis.
• These effects obscure true biological signals, necessitating robust correction methods.

Purpose of the Study:

• To develop an advanced method for correcting batch effects in scRNA-seq data.
• To improve the accuracy and reliability of scRNA-seq data analysis.
• To provide a valuable tool for the biological research community.

Main Methods:

• Designed a dense deep residual network model named NDnetwork.
• Integrated NDnetwork with the MNN method to create the NDMNN method.
• Applied NDMNN to correct batch effects in scRNA-seq datasets.

Main Results:

• NDMNN demonstrated superior performance compared to existing batch effect correction methods.
• Experimental results validated the effectiveness of NDMNN.
• The method successfully mitigates confounding factors in scRNA-seq data.

Conclusions:

• NDMNN is a highly effective tool for correcting batch effects in scRNA-seq data.
• The method enhances the biological insights obtainable from large-scale single-cell studies.
• NDMNN is poised to become an important resource for researchers in the field.