Researchers presented an AI-driven spatial biomarker, ST-DoxPCa, designed to predict which patients with metastatic hormone-sensitive prostate cancer may benefit from adding docetaxel to androgen deprivation therapy (ADT). The work, presented by Sebastian Medina of Emory University School of Medicine at ASCO in Chicago, uses spatial gene-expression signatures inferred from standard H&E pathology slides. The approach aims to address access barriers with genomic assays by using paired spatial transcriptomics and H&E whole-slide images to train a vision transformer model. Medina’s team developed a 26-gene signature tied to androgen receptor signaling and other tumor biology pathways using the HEST-1K prostate cohort. The translational significance is practical: if validated, the method could enable more scalable treatment selection in settings where tissue and genomic testing are constrained by cost, destruction, or infrastructure limits.