DeepPathway: Predicting Pathway Expression from Histopathology Images

Spatial transcriptomics (ST) technologies provide spatially resolved gene expression along with image data, allowing the integrated analysis of complex tissue microenvironments. Despite their potential, the widespread adoption of ST remains limited due to high costs and technical challenges in data acquisition. Thus, there have been recent efforts to develop deep learning (DL)-based computational methods capable of inferring spatial gene expression from H&E images which are much cheaper and easily available. These methods demonstrate promising results in reconstructing transcriptomic landscapes within tissue sections. While existing approaches predominantly focus on gene-level prediction, biological processes are often regulated at the pathway level through coordinated activity among functionally related genes. Here we present DeepPathway, a contrastive learning-based approach trained on ST data to predict pathway expression from H&E images, with input pathway expression computed by summarizing the expression of constituent genes using established gene set definitions. We evaluate the performance of our method on two prostate cancer datasets and validate our approach on the H&E images acquired from The Cancer Genome Atlas (TCGA) clearly differentiating between normal and cancer tissues. Finally, we apply our method to predict hypoxia signatures using H&Es of brain tumour patients where ground truth hypoxia staining was available. Implementation code for DeepPathway is available at https://github.com/aahsan045/DeepPathway .