MAT2: manifold alignment of single-cell transcriptomes with cell triplets.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables detailed cellular analysis.
• Aligning multiple scRNA-seq datasets is crucial for comprehensive biological understanding.
• Existing alignment methods often neglect cell type annotations, limiting robustness.

Purpose of the Study:

• To develop a novel method for aligning single-cell transcriptomes.
• To improve the robustness and accuracy of scRNA-seq data integration.
• To enhance cell type annotation and discover biological variations.

Main Methods:

• Utilized a deep neural network with a contrastive learning strategy.
• Employed cell triplets based on known cell type annotations for alignment.
• Developed MAT2, a method for manifold-based cell alignment.

Main Results:

• MAT2 demonstrated superior performance compared to existing methods on real scRNA-seq datasets.
• The alignment procedure using MAT2 yielded a more robust consensus manifold, especially with limited common cell types.
• Reconstructed gene expression data by MAT2 effectively aided in cell type annotation.
• Identified differential differentiation paces of hematopoietic stem cells between human and mouse.

Conclusions:

• MAT2 offers a robust and effective approach for single-cell transcriptome alignment.
• The method enhances the accuracy of cell type annotation and facilitates biological discovery.
• MAT2 provides a valuable tool for joint analysis of multiple scRNA-seq datasets.