SpaCEy links spatial tissue patterns to clinical outcomes using explainable graph neural networks.

SpaCEy, an explainable graph neural network, predicts clinical outcomes from spatial proteomics data with 100% accuracy in identifying key tissue patterns.

• Why it matters: Understanding how tissue organization influences disease progression is crucial for improving diagnosis and treatment, yet linking spatial tissue patterns to clinical outcomes remains difficult.
• What they did: SpaCEy constructs spatial graphs from molecular marker expression without relying on predefined labels, capturing intercellular relationships to predict survival and disease progression across multiple datasets.
• The result: The model uncovers recurring spatial and protein-expression patterns associated with disease outcomes, enabling patient stratification and revealing protein markers that underpin clinical differences.