RESUMO
Microscopic examination of prostate cancer has failed to reveal a reproducible association between molecular and morphologic features. However, deep-learning algorithms trained on hematoxylin and eosin (H&E)-stained whole slide images (WSI) may outperform the human eye and help to screen for clinically-relevant genomic alterations. We created deep-learning algorithms to identify prostate tumors with underlying ETS-related gene (ERG) fusions or PTEN deletions using the following 4 stages: (1) automated tumor identification, (2) feature representation learning, (3) classification, and (4) explainability map generation. A novel transformer-based hierarchical architecture was trained on a single representative WSI of the dominant tumor nodule from a radical prostatectomy (RP) cohort with known ERG/PTEN status (n = 224 and n = 205, respectively). Two distinct vision transformer-based networks were used for feature extraction, and a distinct transformer-based model was used for classification. The ERG algorithm performance was validated across 3 RP cohorts, including 64 WSI from the pretraining cohort (AUC, 0.91) and 248 and 375 WSI from 2 independent RP cohorts (AUC, 0.86 and 0.89, respectively). In addition, we tested the ERG algorithm performance in 2 needle biopsy cohorts comprised of 179 and 148 WSI (AUC, 0.78 and 0.80, respectively). Focusing on cases with homogeneous (clonal) PTEN status, PTEN algorithm performance was assessed using 50 WSI reserved from the pretraining cohort (AUC, 0.81), 201 and 337 WSI from 2 independent RP cohorts (AUC, 0.72 and 0.80, respectively), and 151 WSI from a needle biopsy cohort (AUC, 0.75). For explainability, the PTEN algorithm was also applied to 19 WSI with heterogeneous (subclonal) PTEN loss, where the percentage tumor area with predicted PTEN loss correlated with that based on immunohistochemistry (r = 0.58, P = .0097). These deep-learning algorithms to predict ERG/PTEN status prove that H&E images can be used to screen for underlying genomic alterations in prostate cancer.
RESUMO
BACKGROUND: The need to better understand the molecular underpinnings of the heterogeneous outcomes of patients with prostate cancer is a pressing global problem and a key research priority for Movember. To address this, the Movember Global Action Plan 1 Unique tissue microarray (GAP1-UTMA) project constructed a set of unique and richly annotated tissue microarrays (TMA) from prostate cancer samples obtained from multiple institutions across several global locations. METHODS: Three separate TMA sets were built that differ by purpose and disease state. RESULTS: The intended use of TMA1 (Primary Matched LN) is to validate biomarkers that help determine which clinically localized prostate cancers with associated lymph node metastasis have a high risk of progression to lethal castration-resistant metastatic disease, and to compare molecular properties of high-risk index lesions within the prostate to regional lymph node metastases resected at the time of prostatectomy. TMA2 (Pre vs. Post ADT) was designed to address questions regarding risk of castration-resistant prostate cancer (CRPC) and response to suppression of the androgen receptor/androgen axis, and characterization of the castration-resistant phenotype. TMA3 (CRPC Met Heterogeneity)'s intended use is to assess the heterogeneity of molecular markers across different anatomic sites in lethal prostate cancer metastases. CONCLUSIONS: The GAP1-UTMA project has succeeded in combining a large set of tissue specimens from 501 patients with prostate cancer with rich clinical annotation. IMPACT: This resource is now available to the prostate cancer community as a tool for biomarker validation to address important unanswered clinical questions around disease progression and response to treatment.
