A new spatial transcriptomics–guided biomarker approach presented at ASCO aimed to predict which metastatic hormone-sensitive prostate cancer patients are more likely to benefit from adding docetaxel to androgen deprivation therapy. The model, ST-DoxPCa, was presented by Sebastian Medina, a PhD candidate at Emory University School of Medicine, using AI inference from H&E pathology slides. The work focused on building a prostate cancer-specific gene signature using paired spatial transcriptomics and H&E images. Medina’s team trained a vision transformer using publicly available cases from the HEST-1K cohort and produced a 26-gene signature tied to androgen receptor signaling, oncogenesis, tumor suppression, stemness, metabolic processes, and inflammation. The stated goal is to provide a more accessible alternative to genomic assays, which can be costly, tissue-destructive, and difficult to deploy in resource-limited settings. The approach is positioned as potentially scalable because it leverages standard pathology workflows. If validated in prospective studies, the strategy could reshape patient stratification for docetaxel in metastatic hormone-sensitive disease by lowering assay barriers while preserving predictive performance.