Discriminative Domain Adaption Network for Simultaneously Removing Batch Effects and Annotating Cell Types in...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular heterogeneity.
• Batch effects in scRNA-seq data hinder accurate cell type identification and analysis.
• Existing methods often overlook global distribution matching and discriminative features in batch correction.

Purpose of the Study:

• To develop a novel method for joint batch effect correction and cell type annotation in scRNA-seq data.
• To address limitations of current batch correction algorithms by incorporating global distribution matching and discriminative feature learning.
• To enhance the robustness and accuracy of scRNA-seq analysis.

Main Methods:

• Proposed the discriminative domain adaption network (D2AN) for scRNA-seq data.
• Employed adversarial domain adaptation to capture global low-dimensional embeddings.
• Utilized contrastive loss and semantic alignment of class centroids for local alignment.
• Implemented a self-paced learning mechanism for robust model training.

Main Results:

• D2AN effectively corrects batch effects in scRNA-seq data.
• The method achieves accurate cell type annotation by aligning global and local data distributions.
• Experimental results show superior performance compared to state-of-the-art methods on multiple real datasets.

Conclusions:

• D2AN offers a powerful approach for integrated batch correction and cell type annotation in scRNA-seq.
• The proposed method enhances the reliability of scRNA-seq data analysis for biological discovery.
• This work advances the field of computational single-cell genomics.