Learning Cell Annotation under Multiple Reference Datasets by Multisource Domain Adaptation.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate cell type annotation is crucial for single-cell sequencing analysis.
• Current methods often rely on single reference datasets, limiting performance with increasing data complexity.
• Integrating multiple reference datasets is challenging due to batch effects.

Purpose of the Study:

• To develop a novel method for cell type annotation that integrates multiple reference datasets.
• To address the challenge of batch effects in multi-dataset integration for single-cell analysis.
• To improve the accuracy and efficiency of cell type assignment in single-cell sequencing data.

Main Methods:

• Proposed scMDR and scMultiR, utilizing multisource domain adaptation.
• Learned cell type-specific information from multiple reference datasets and query cells.
• Developed algorithms to overcome batch effects for integrative analysis.

Main Results:

• scMDR and scMultiR demonstrated state-of-the-art effectiveness.
• The methods successfully integrated multiple reference datasets for improved annotation.
• Benchmark experiments validated the performance of the proposed approaches.

Conclusions:

• scMDR and scMultiR offer a powerful solution for integrative single-cell assignment.
• The developed methods enhance cell type annotation by leveraging multisource domain adaptation.
• This work advances the field of single-cell data analysis by addressing multi-dataset integration challenges.