Decoding Spatial Heterogeneity and Multi-Omics Regulation with Hierarchical Graph Learning.

Area of Science:

• Spatial biology
• Multi-omics analysis
• Computational biology

Background:

• Spatial multi-omics technologies enable simultaneous profiling of multiple molecular layers in tissues.
• Existing computational methods often lack interpretability, failing to uncover underlying regulatory mechanisms of spatial heterogeneity.

Purpose of the Study:

• Introduce SMOReg, a novel spatial multi-omics hierarchical graph learning framework.
• Enhance biological interpretability of spatial domains by embedding cross-omics regulatory patterns.

Main Methods:

• Developed SMOReg, a hierarchical graph learning framework.
• Employed cross-graph matching to integrate cross-omics regulatory patterns into spot representations.
• Evaluated on paired spatial transcriptomic and proteomic datasets.

Main Results:

• SMOReg outperformed existing methods across multiple metrics in identifying spatial domains.
• Uncovered fine-grained spatial domains with interpretable multi-layer regulatory signatures.
• Successfully distinguished germinal center light and dark zones in human tonsil, revealing specialized regulatory pathways.

Conclusions:

• SMOReg provides a powerful framework for deciphering spatial domain heterogeneity and multi-omics regulation.
• Significantly advances the interpretability of spatial domains in complex tissues.
• Enables deeper understanding of tissue organization and cellular functions.