Identification of disease-relevant cellular microenvironments is a critical step to deriving clinically actionable insights from spatial proteomics data. Spatial signatures, including the cell type composition of these microenvironments and the local spatial relationships between cells within them, can aid in understanding mechanism of disease or drug response, as well as patient stratification. To address this, scientists at Enable Medicine and collaborators developed SPAtial CEllular Graphical Modeling (SPACE-GM), a geometric deep learning framework to predict clinical phenotypes from cellular microenvironments and to further characterize those that are disease-relevant.

‍

This insight report will provide a characterization of the cellular microenvironments identified by SPACE-GM that correlate with patient clinical outcomes.