Significance: Renal cell carcinomas (RCCs) are heterogeneous malignancies thought to arise from kidney tubular epithelial cells, and clear cell RCC is the most common entity. This study demonstrates that cell atlases generated from benign kidney and two common RCCs using single-cell RNA sequencing can predict putative cells of origin for more than 10 RCC subtypes. A focused analysis of distinct cell-type compartments reveals the potential role of tumor epithelia in promoting immune infiltration and other molecular attributes of the tumor microenvironment. Finally, an observed association between the lack of immunotherapy response and endothelial cell fraction has important clinical implications. The current study, therefore, significantly contributes toward understanding disease ontogenies and the molecular dynamics of tumor epithelia and the microenvironment.

Prostate cancer is the most frequent cancer in men and a leading cause of cancer death. Determining a patient’s optimal therapy is a challenge, where oncologists must select a therapy with the highest likelihood of success and the lowest likelihood of toxicity. International standards for prognostication rely on non-specific and semi-quantitative tools, commonly leading to over- and under-treatment. Tissue-based molecular biomarkers have attempted to address this, but most have limited validation in prospective randomized trials and expensive processing costs, posing substantial barriers to widespread adoption. There remains a significant need for accurate and scalable tools to support therapy personalization. Here we demonstrate prostate cancer therapy personalization by predicting long-term, clinically relevant outcomes using a multimodal deep learning architecture and train models using clinical data and digital histopathology from prostate biopsies. We train and validate models using five phase III randomized trials conducted across hundreds of clinical centers. Histopathological data was available for 5654 of 7764 randomized patients (71%) with a median follow-up of 11.4 years. Compared to the most common risk-stratification tool—risk groups developed by the National Cancer Center Network (NCCN)—our models have superior discriminatory performance across all endpoints, ranging from 9.2% to 14.6% relative improvement in a held-out validation set. This artificial intelligence-based tool improves prognostication over standard tools and allows oncologists to computationally predict the likeliest outcomes of specific patients to determine optimal treatment. Outfitted with digital scanners and internet access, any clinic could offer such capabilities, enabling global access to therapy personalization.

BACKGROUND. Commercial gene expression assays are guiding clinical decision making in patients with prostate cancer, particularly when considering active surveillance. Given heterogeneity and multifocality of primary prostate cancer, such assays should ideally be robust to the coexistence of unsampled higher grade disease elsewhere in the prostate in order to have clinical utility. Herein, we comprehensively evaluated transcriptomic profiles of primary multifocal prostate cancer to assess robustness to clinically relevant multifocality. METHODS. We designed a comprehensive, multiplexed targeted RNA-sequencing assay capable of assessing multiple transcriptional classes and deriving commercially available prognostic signatures, including the Myriad Prolaris Cell Cycle Progression score, the Oncotype DX Genomic Prostate Score, and the GenomeDX Decipher Genomic Classifier. We applied this assay to a retrospective, multi-institutional cohort of 156 prostate cancer samples. Derived commercial biomarker scores for 120 informative primary prostate cancer samples from 44 cases were determined and compared. RESULTS. Derived expression scores were positively correlated with tumor grade ( r _S = 0.53–0.73; all P < 0.001), both within the same case and across the entire cohort. In cases of extreme grade-discordant multifocality (co-occurrence of grade group 1 [GG1] and ≥GG4 foci], gene expression scores were significantly lower in low- (GG1) versus high-grade (≥GG4) foci (all P < 0.001). No significant differences in expression scores, however, were observed between GG1 foci from prostates with and without coexisting higher grade cancer (all P > 0.05). CONCLUSIONS. Multifocal, low-grade and high-grade prostate cancer foci exhibit distinct prognostic expression signatures. These findings demonstrate that prognostic RNA expression assays performed on low-grade prostate cancer biopsy tissue may not provide meaningful information on the presence of coexisting unsampled aggressive disease. FUNDING. Prostate Cancer Foundation, National Institutes of Health (U01 CA214170, R01 CA183857, University of Michigan Prostate Specialized Program of Research Excellence [S.P.O.R.E.] P50 CA186786-05, Weill Cornell Medicine S.P.O.R.E. P50 CA211024-01A1), Men of Michigan Prostate Cancer Research Fund, University of Michigan Comprehensive Cancer Center core grant (2-P30-CA-046592-24), A. Alfred Taubman Biomedical Research Institute, and Department of Defense.

Metastatic castration-resistant prostate cancer (mCRPC) encompasses a heterogeneous wide range of molecular tumor behavior and a high risk of progression. Early detection and treatment are therefore crucial in these patients. Treatment has improved drastically in recent years and many novel therapeutic agents are currently under investigation. However, due to the rapidly changing therapeutic landscape in mCRPC, it is difficult for clinicians to keep up to date with the latest innovations in this area. In the present narrative review, we discuss the current and emerging therapies for mCRPC as well as the clinical and molecular factors that can help predict which patients are most likely to benefit from these novel agents.