A spatial-transcriptomics approach presented at ASCO predicted docetaxel benefit in metastatic hormone-sensitive prostate cancer using only routine H&E slides as a starting point. The biomarker model, ST-DoxPCa, was built with a vision transformer and trained on paired spatial transcriptomics and histology from the HEST-1K cohort. The investigators described a 26-gene signature related to androgen receptor signaling and other cancer biology pathways, aiming to deliver treatment stratification where genomic testing is costly or impractical. The work also frames an implementation pathway for scaling biomarker selection in lower-resource settings by leveraging AI-derived spatial inference rather than tissue-destructive genomic profiling. If replicated, it could broaden access to precision oncology selection and accelerate adoption for clinicians looking to use standardized slide workflows.