PROJECT SUMMARY: There were >34,000 PCa-related deaths in 2020 in the US alone. Definitive treatment includes Radical prostatectomy (RP) or radiotherapy (RT) with long term androgen-suppression therapy (ADT). These have been shown to be effective treatments for organ-confined PCa and have been demonstrated to reduce the risk of death from PCa. In 38-52% of cases, however, advanced disease with potentially poor prognosis is found on tissue pathology. A number of recent clinical trials have shown the benefit of adjuvant therapy in select PCa patients post-RP or RT. However, it is critical to identify those PCa patients who following definitive therapy (surgery or radiation) are at high-risk for recurrence or metastasis and thus will benefit from adjuvant therapy versus patients who will not and hence may be spared the morbidity and cost of therapy. Recognizing the significance of this unmet clinical need, in 2018 the NCCN guidelines for PCa were modified to include the Decipher Score, a prognostic molecular gene-based test to identify the likelihood of metastasis following surgery. We have developed our own "Integrated Risk Score" (IRiS) image classifier that (npj Precison Onc, In Press14) combines computer extracted morphologic glandular features from H&E tissue slides of the tumor. IRiS stratified PCa patients (N>900, 6 sites) based on their time to biochemical recurrence (BCR) into low- and high-risk groups (p<0.001; HR=2.44). Further, IRiS when combined with pre-op PSA and Gleason grade outperformed Decipher in predicting BCR in N=173 patients (p<0.001; HR=3.23 vs HR=2.76). In this R01, we will validate IRiS as (1) prognostic of BCR and risk of metastasis as well as (2) predictive of the added benefit of additional chemotherapy following definitive therapy (surgery or radiation) in PCa. In a recent paper in Clin Cancer Res, we identified IRiS specific prognostic features for African American (AA) men with PCa. We will build on these findings to develop population specific IRiS models for PCa. We will also further optimize IRiS by including (1) features of stromal and cribriform morphology, (2) develop population specific IRiS models for different ethnic groups, and (3) complement IRiS with clinico-pathological features. To validate IRiS as predictive of benefit of adjuvant therapy, we need access to randomized clinical trial tissue slide images involving PCa patients treated with definitive therapy alone (surgery or ADT+radiation) and definitive therapy+ adj. chemo. The STAMPEDE and RTOG-0521 trials fit these criteria; we have secured approval to access tissue slide images from these trials. To make the tool widely available, IRiS will be integrated into PathPresenter, a digital pathology viewer and management platform currently in use in 178 countries. This partnership will combine expertise in (a) computational pathology of the Madabhushi group, (2) clinical, pathological and biomarker expertise of PCa from the University of Pennsylvania (D...