scMoMaT jointly performs single cell mosaic integration and multi-modal bio-marker detection

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell data integration is crucial for combining information across batches and modalities.
• Mosaic integration, the most general scenario, remains challenging with limited available methods.

Purpose of the Study:

• To introduce scMoMaT, a new method for mosaic single-cell multi-omics data integration.
• To address the challenge of integrating datasets with unequal cell type compositions.

Main Methods:

• scMoMaT utilizes matrix tri-factorization for integrating single-cell multi-omics data.
• The method identifies cluster-specific biomarkers across different modalities.

Main Results:

• scMoMaT successfully integrates single-cell multi-omics data in the mosaic scenario.
• It accurately annotates cell types using multi-modal biomarkers, improving upon existing methods.
• The method demonstrates superior performance on real and simulated datasets, handling unequal cell type compositions.

Conclusions:

• scMoMaT provides a robust solution for mosaic single-cell multi-omics integration.
• The integrated cell embeddings and learned biomarkers enhance cell type annotation quality and resolution.