The integration of routine histopathology with advanced spatial omics technologies is transforming precision medicine. Haematoxylin and eosin (H&E) stained images, traditionally used for morphological assessment, contain molecular signals that can now be computationally decoded to predict protein and gene expression patterns, normally reliant on costly multiplex assays. Spatial transcriptomics preserves the spatial context of gene expression, while AI-driven imaging extracts detailed tissue morphology. Combining these approaches offers a holistic view of tissue architecture and disease biology.
This project leverages the King’s Health Partner (KHP) Cancer Biobank, which provides tissue samples matched with longitudinal clinical records and patient outcomes, enabling robust clinical validation. By linking computational predictions from H&E-stained tissues to real-world patient trajectories, the research ensures that models are clinically relevant and actionable.
In parallel, patient-derived organoid (PDO) models offer an independent platform for functional validation. PDOs replicate tumour complexity and allow controlled drug-response experiments, creating opportunities to test therapeutic hypotheses and uncover mechanisms of resistance. This dual strategy, clinical outcome validation through tissue samples and functional validation through PDOs, bridges computational pathology with translational medicine.
The work focuses on developing disease-specific AI models, refining spatial omics workflows, and prioritising clinically actionable biomarkers. Multi-layered validation (molecular, functional, and clinical) ensures predictions are biologically meaningful and generalisable. Ultimately, this approach reduces costs, accelerates molecular characterisation, and unlocks retrospective analysis of archival H&E-stained tissue samples. By combining computational insights with experimental systems, the project advances personalised treatment strategies and functional precision medicine.