scRDAN: a robust domain adaptation network for cell type annotation across single-cell RNA sequencing data.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) is vital for understanding cellular heterogeneity.
• Accurate cell type annotation in scRNA-seq data is challenged by noise and batch effects.
• Existing methods struggle to robustly handle these data complexities.

Purpose of the Study:

• To introduce scRDAN, a novel robust domain adaptation network for accurate cell type annotation in scRNA-seq data.
• To address noise interference and batch effects inherent in scRNA-seq analysis.
• To enhance the robustness and generalization capabilities of cell annotation models.

Main Methods:

• scRDAN employs three modules: denoising domain adaptation, fine-grained discrimination, and robustness enhancement.
• Denoising domain adaptation uses feature reconstruction and adversarial learning to align data distributions.
• Fine-grained discrimination improves cell type distinction by enhancing separability and compactness.
• Robustness enhancement introduces noise to improve model generalization.

Main Results:

• scRDAN demonstrated superior performance in handling batch effects compared to existing methods.
• The method significantly improved cell type annotation accuracy across diverse datasets.
• Evaluations on simulated, cross-platform, and cross-species data confirmed scRDAN's effectiveness.

Conclusions:

• scRDAN offers a robust solution for accurate cell type annotation in scRNA-seq data.
• The proposed network effectively mitigates noise and batch effects, enhancing analytical reliability.
• scRDAN represents a significant advancement in computational tools for single-cell analysis.