DeepPathway: Predicting Pathway Expression from Histopathology Images.

MOTIVATION: Spatial transcriptomics (ST) technologies provide spatially resolved gene expression along with image data, allowing the integrative analysis of complex tissue microenvironment. Despite their potential, the widespread adoption of ST remains limited due to high costs, and methodological challenges in data acquisition. Thus, there have been recent efforts to develop deep learning methods for inferring spatial gene expression from much cheaper and easily available haematoxylin and eosin (H&E) images. These methods demonstrate promising results in reconstructing transcriptomic landscapes within tissue sections. While existing approaches focus on gene-level predictions, biological processes are often regulated at the pathway level through coordinated activity among functionally related genes. RESULTS: We present DeepPathway, a contrastive learning-based approach trained on ST data to predict pathway expression from H&Es. We compute input pathway expression by summarizing the expression of constituent genes using established pathway definitions. We evaluate the performance of DeepPathway on multiple cancer datasets and validate it on the H&E images from The Cancer Genome Atlas (TCGA) clearly differentiating certain pathway activities in normal and tumour tissue regions. Finally, we apply our method to predict hypoxia signatures using H&Es of brain tumour samples where hypoxia staining with pimonidazole was available as ground truth. CODE AVAILABILITY: Implementation code for DeepPathway is available at https://doi.org/10.5281/zenodo.21100191 and at GitHub repository: https://github.com/aahsan045/DeepPathway.