Dissecting spatial patterning and signaling with directional diffusion in spatial multi-omics.

Area of Science:

• Computational biology
• Genomics
• Systems biology

Background:

• Spatial multi-omics sequencing offers high-resolution insights into tissue organization and molecular regulation.
• Integrating multiple omics modalities within a spatial context presents significant analytical challenges.

Purpose of the Study:

• To introduce SpaDDM, a novel framework for spatial multi-omics integration using directional diffusion models (DDMs).
• To enable spatial pattern identification, cross-omics alignment, and analysis of intercellular signaling flows.

Main Methods:

• SpaDDM utilizes DDM-based graph networks to learn omics-specific representations incorporating spatial and molecular data.
• An attention mechanism is employed for feature alignment across different omics modalities.
• The framework was benchmarked on diverse spatial multi-omics datasets (transcriptomics-epigenomics, transcriptomics-proteomics).

Main Results:

• SpaDDM accurately identified spatial tissue patterns and reduced boundary noise between regions, outperforming existing methods.
• The framework effectively aligned features across multiple omics layers, facilitating cross-omics translation.
• Learned low-dimensional cell representations enabled inference of signaling pathways underlying spatial patterning.

Conclusions:

• SpaDDM provides a robust approach for spatial multi-omics integration and analysis.
• The framework enhances understanding of tissue organization, cellular communication, and cell state prediction.
• SpaDDM facilitates the translation of complementary information across omics modalities.