Characterizing spatial functional microniches with SpaceTravLR.

Area of Science:

• Spatial biology
• Computational biology
• Machine learning in genomics

Background:

• Spatial omics technologies offer insights into tissue biology but lack functional prediction capabilities.
• Current in-silico perturbation methods lack spatial resolution, limiting their application to single-cell data.
• Existing methods identify morphological domains but not spatially resolved functional microniches.

Purpose of the Study:

• To develop an interpretable machine learning approach, SpaceTravLR, for uncovering spatial functional microniches.
• To model the impact of genetic perturbations on tissue organization and function with spatial resolution.
• To generate experimentally validated mechanistic hypotheses for identified spatial niches.

Main Methods:

• SpaceTravLR uses spatial omics data to infer how genetic perturbations rewire signals across tissue neighborhoods.
• The approach propagates perturbation effects through spatially resolved molecular networks.
• It generalizes across tissues and species to identify novel spatial microniches.

Main Results:

• SpaceTravLR successfully defines novel spatial microniches across various tissues and contexts.
• Perturbation predictions align with experimental validation and known mechanistic outcomes.
• A novel mechanism for Ccr4 in T helper 2 cell localization was discovered and validated in a murine model.

Conclusions:

• SpaceTravLR provides a validated framework for understanding gene function in spatial tissue organization.
• The tool uncovers how genes act individually and combinatorially to shape tissue function.
• It enables the generation of mechanistic hypotheses for spatial niche identification.