scMFG: a single-cell multi-omics integration method based on feature grouping.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Simultaneous measurement of multi-omics data offers comprehensive cellular heterogeneity insights.
• Existing methods struggle with accurate cell type identification and interpretability, especially with noisy data.

Purpose of the Study:

• To introduce scMFG, a novel method for integrating single-cell multi-omics data.
• To address limitations in cell type identification and model interpretability in existing methods.

Main Methods:

• scMFG utilizes feature grouping and group integration for single-cell multi-omics data.
• It organizes features within omics layers and ensures consistent grouping across layers.
• A matrix factorization approach enhances the interpretability of integrated results.

Main Results:

• scMFG demonstrated robust cell type identification across four real-world datasets.
• It achieved superior performance in resolving cellular heterogeneity at finer resolutions on simulated data.
• The method effectively identified rare cell types and validated interpretability with specific cell states.
• scMFG identified cell developmental trajectories, even in datasets with batch effects.

Conclusions:

• scMFG provides a robust framework for analyzing single-cell multi-omics data.
• It advances the understanding of cellular heterogeneity in a comprehensive and interpretable way.