Jia-Shu He

Learning Interpretable Tumor Microenvironment Representations by Fitting Pan-Cancer Cell State-Niche Correlation

GITIII-scale is presented, a hierarchical, interpretable pan-cancer spatial transcriptomics foundation model for TME representation learning that investigates cell state-niche associations and their underlying ligand-receptor (LR) signaling pathways that recovered niche-associated state changes more accurately than existing spatial transcriptomics foundation models in cancer types unseen during training.