Interpretable data integration for single-cell and spatial multi-omics.

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Single-cell and spatial transcriptomic data integration with epigenomic data is crucial for understanding cell fate.
• Existing methods struggle to link genes with regulatory elements and dissect cell-specific regulation.
• Current approaches often align data into a shared latent space, losing specific regulatory connections.

Purpose of the Study:

• To develop a novel method for integrating multi-omics data that simultaneously identifies cellular heterogeneity and transcriptional regulatory links.
• To provide post hoc interpretability for existing data integration techniques.
• To enable the dissection of cell-state and spatial-location-specific transcriptional regulation.

Main Methods:

• Feature-guided optimal transport (FGOT) method was developed for multi-omics data integration.
• FGOT handles paired/unpaired single-cell and paired spatial multi-omics data.
• Method validated using histone modification and 3D genomics data.

Main Results:

• FGOT accurately integrates multi-omics data and infers regulatory links.
• The method demonstrates robustness and accuracy in uncovering cellular heterogeneity.
• FGOT successfully identifies cell-state and spatial-location-specific regulatory elements.

Conclusions:

• FGOT offers a powerful approach for dissecting transcriptional regulation across diverse cellular states and spatial contexts.
• The method enhances the interpretability of multi-omics data integration for disease mechanism studies.
• FGOT facilitates systematic screening of regulatory elements at the single-cell level.