scAlign: a tool for alignment, integration, and rare cell identification from scRNA-seq data.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) data integration is crucial for comparative analyses across different biological conditions, species, or experimental batches.
• Existing methods face challenges in handling variations in cell type composition and gene expression patterns between datasets.

Purpose of the Study:

• To introduce scAlign, a novel unsupervised deep learning framework for robust scRNA-seq data integration.
• To enable accurate estimation of per-cell gene expression differences across datasets, accommodating varying levels of cell label information.

Main Methods:

• scAlign employs an unsupervised deep learning approach for data integration.
• The method can utilize partial, overlapping, or complete cell labels.
• It estimates per-cell gene expression differences and is robust to cross-dataset variations.

Main Results:

• scAlign achieves state-of-the-art performance in scRNA-seq data integration.
• The method demonstrates robustness against variations in cell type-specific expression and composition.
• scAlign successfully identified gene expression programs in rare malaria parasite populations.

Conclusions:

• scAlign provides a powerful and flexible tool for scRNA-seq data integration.
• The framework is effective in identifying biological insights, even in rare cell populations.
• Its applicability extends to diverse data integration challenges across various scientific domains.