Prostate cancer encompasses a spectrum of disease states driven by complex cellular heterogeneity. To delineate the transcriptional programs underlying lineage plasticity and metastasis, we constructed a comprehensive single-cell atlas of 128 patients, spanning localized, castration-resistant, and metastatic disease. Lineage plasticity was prevalent in localized disease, with subsets of tumor cells adopting distinct basal-like and club-like states. Luminal-like cancer cells also displayed extensive lineage infidelity, defined not by a binary loss of identity but by the combinatorial erosion of luminal gene modules associated with higher grade and stage. In the metastatic setting, gene program association analysis (GPAS) identified a broad induction of cell-cycle gene modules across organ sites as well as an induction of organ-specific gene modules, including osteomimetic signaling in bone, neuro-migratory genes in brain, and erythroid-like transitions in liver. Neuroendocrine prostate cancers (NEPCs) were not monolithic but defined by combinations of NE-associated gene modules including a novel HES6 program. Notably, these modules were detected at intermediate levels in localized samples, suggesting molecular plasticity precedes histological transformation. We also developed a refined NE signature that could distinguish NEPC tumors more accurately than previously published signatures. Within the tumor microenvironment (TME), we observed an elevation of pro-inflammatory Th17 T-cells in African American patients and identified a rare Schwann cell population. Finally, we present PCformer, a transformer-based foundation model trained on >500,000 cells to automate cell-state classification. Together, this comprehensive atlas demonstrates the complex nature of gene modules underlying lineage infidelity and plasticity in cancer cells and highlights distinct immune and stromal populations within the tumor ecosystem.

Abstract Introduction: The potential for long-term exercise to affect the metabolome in healthy individuals is well established. Given the metabolic underpinnings of prostate cancer, it is important to investigate whether long-term exercise similarly alters the metabolome in disease-affected men. This study is among the first to characterize metabolic changes in individuals with prostate cancer, comparing those assigned to a home-based walking program to those assigned to printed materials with physical activity recommendations. Methods: Fifty-one men with prostate cancer on active surveillance were randomly allocated to the exercise or control intervention. Metabolomic profiling of primary metabolism, complex lipids, and biogenic amines was performed at the West Coast Metabolomics Center on serum samples collected at baseline and after the 16-week interventions; data were successfully generated for 22 participants in the exercise arm and 23 participants in the control arm. To identify intervention-related differences in metabolic changes, we performed hierarchical clustering and fit mixed-effects models including an arm x time interaction. Results: Hierarchical clustering indicated limited separation in metabolic profiles between the exercise and control groups at 16 weeks. Although none of the 1,220 named metabolites exhibited statistically significant differences in change between the two groups (q<0.10), 85 (7.0%) demonstrated nominal significance (p<0.05). Among the 15 metabolites with the smallest p-values, six (40%) were sphingolipids - specifically sphingomyelins - though sphingolipids comprised only 11% of all named metabolites. All six sphingomyelins decreased more over time in the exercise group than in the control group. Conclusions: Although metabolic profiles were not significantly altered overall, a walking intervention may promote the reduction of sphingomyelins, thereby shifting lipid signaling toward pathways that enhance mitochondrial function and reduce inflammation. Such changes are consistent with biologically plausible mechanisms through which exercise could favorably influence prostate cancer biology, even in the absence of broad metabolomic shifts. Citation Format: Rebecca E. Graff, Ritu Roy, Oliver Fiehn, Adam Olshen, Erin Van Blarigan, Stacey Kenfield, Jeffry P. Simko, Anthony Luke, Lee Jones, Matthew R. Cooperberg, Peter R. Carroll, June M. Chan. Effects of a home-based walking intervention on serum metabolomic profiles in men with prostate cancer on active surveillance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1237.

Prostate cancer encompasses a spectrum of disease states driven by complex cellular heterogeneity. To delineate the transcriptional programs underlying lineage plasticity and metastasis, we constructed a comprehensive single-cell atlas of 128 patients, spanning localized, castration-resistant, and metastatic disease. Lineage plasticity was prevalent in localized disease, with subsets of tumor cells adopting distinct basal-like and club-like states. Luminal-like cancer cells also displayed extensive lineage infidelity, defined not by a binary loss of identity but by the combinatorial erosion of luminal gene modules associated with higher grade and stage. In the metastatic setting, gene program association analysis (GPAS) identified a broad induction of cell-cycle gene modules across organ sites as well as an induction of organ-specific gene modules, including osteomimetic signaling in bone, neuro-migratory genes in brain, and erythroid-like transitions in liver. Neuroendocrine prostate cancers (NEPCs) were not monolithic but defined by combinations of NE-associated gene modules including a novel HES6 program. Notably, these modules were detected at intermediate levels in localized samples, suggesting molecular plasticity precedes histological transformation. We also developed a refined NE signature that could distinguish NEPC tumors more accurately than previously published signatures. Within the tumor microenvironment (TME), we observed an elevation of pro-inflammatory Th17 T-cells in African American patients and identified a rare Schwann cell population. Finally, we present PCformer, a transformer-based foundation model trained on >500,000 cells to automate cell-state classification. Together, this comprehensive atlas demonstrates the complex nature of gene modules underlying lineage infidelity and plasticity in cancer cells and highlights distinct immune and stromal populations within the tumor ecosystem.

PURPOSE Radiology reports are stored as plain text in most electronic health records, rendering the data computationally inaccessible. Large language models are powerful tools for analyzing unstructured text but relatively untested in urologic oncology. We aimed to develop a pipeline to extract data from plain text prostate magnetic resonance imaging (MRI) reports using GPT4.0 and compare the accuracy to manually abstracted data. METHODS We developed a data pipeline using a secure, enterprise-wide deployment of OpenAI's GPT-4.0 to automatically extract data elements from prostate MRI report text when presented with prostate MRI reports. Identical prompts and reports were sent multiple times to determine response variability. We extracted 15 data elements per report and compared accuracy to a manually abstracted gold standard. RESULTS Across 424 prostate MRI reports, GPT-4.0 response accuracy was consistently above 95%. Individual field accuracies were 98.3% (96.3%-99.3%) for prostate-specific antigen density, 97.4% (95.4%-98.7%) for extracapsular extension, and 98.1% (96.3%-99.2%) for TNM stage, and had a median of 98.1% (96.3%-99.2%), a mean of 97.2% (95.2%-98.3%), and a range of 99.8% (98.7%-100.0%) for number of suspicious lesions to 87.7% (84.2%-90.7%) for identification of lesion location in the base of the prostate. Response variability over five repeated runs ranged from 0.14% to 3.61%, differed based on the data element extracted (P < .001), and was inversely correlated with accuracy (P < .001). In disagreements between manual and GPT-4.0 extracted data, GPT-4.0 responses were more often deemed correct by an additional reviewer. CONCLUSION GPT-4.0 had high accuracy with low variability in extracting data points from prostate cancer MRI reports with low upfront programming requirements. This represents an effective tool to expedite medical data extraction for clinical and research use cases.

ObjectivesProstate cancer is the most common cancer among men in the United States. This study examines factors associated with active surveillance (AS) uptake, timing of treatment decision making, and whether timing affects quality of life.MethodsWe used data from a population-based observational study of 512 patients aged 40–79 diagnosed with low-risk prostate cancer from 2016 to 2022. Factors associated with AS receipt and treatment decision-making trajectories were assessed using robust Poisson regression models. Patient-reported satisfaction with treatment and PROMIS measures were analyzed using Poisson and linear regression.ResultsAS uptake was 70.9% in the overall sample, higher among non-Hispanic (NH) White (76.6%) and NH Asian American/Pacific Islander (75.0%) patients than among NH Black (64.1%) and Hispanic (57.3%) patients. In adjusted models, higher education was associated with greater AS uptake (prevalence ratio (PR): 1.35; 95% CI: 1.02–1.78), and Hispanic patients were less likely to choose AS (PR: 0.80; 95% CI: 0.65–0.99). Older patients were more likely to delay disease management decisions (60–69 years: PR: 1.47; 95% CI: 1.09–1.98; 70 + years: PR: 1.67; 95% CI: 1.17–2.37), while partnered patients were less likely to delay (PR: 0.69; 95% CI: 0.54–0.88). Late deciders reported lower satisfaction with treatment decisions (PR: 0.92; 95% CI: 0.85–0.99) and higher anxiety scores at follow-up (coef: 1.86; 95% CI: 0.15–3.57).ConclusionsHispanic patients, those who completed less education, and older patients face barriers to timely decisions about disease management. Partner involvement supports earlier decisions, while later decision was linked to lower satisfaction and higher anxiety.

Physical activity is associated with reduced risk of prostate cancer (PCa) progression and death; changes in insulin sensitivity and inflammation are potential mediating mechanisms. This study examined whether exercise after PCa diagnosis affects insulin-related and inflammatory biomarkers. The Active Surveillance Exercise (ASX) randomized controlled trial was assigned to men undergoing active surveillance for low-risk, localized PCa to a 16-week exercise intervention (home-based walking program; n = 26) or printed physical activity recommendations (control group; n = 25). Fasting blood samples were collected at baseline and after 16 weeks. Samples were analyzed for markers of insulinemia (insulin, C-peptide, adiponectin), inflammation (C-reactive protein (CRP)), and prostate-specific antigen (PSA). Biomarker changes over time and between arms were analyzed using linear mixed-effects models and intention-to-treat analysis. 22 (85

Background:The manual abstraction of unstructured clinical data is often necessary for granular clinical outcomes research but is time consuming and can be of variable quality. Large language models (LLMs) show promise in medical data extraction yet integrating them into research workflows remains challenging and poorly described. Objective:This study aimed to develop and integrate an LLM-based system for automated data extraction from unstructured electronic health record (EHR) text reports within an established clinical outcomes database. Methods:We implemented a generative artificial intelligence pipeline (UODBLLM) utilizing a flexible language model interface that supports various LLM implementations, including Health Insurance Portability and Accountability Act-compliant cloud services and local open-source models. We used extensible markup language (XML)-structured prompts and integrated using an open database connectivity interface to generate structured data from clinical documentation in the EHR. We evaluated the UODBLLM's performance on the completion rate, processing time, and extraction capabilities across multiple clinical data elements, including quantitative measurements, categorical assessments, and anatomical descriptions, using sample magnetic resonance imaging (MRI) reports as test cases. System reliability was tested across multiple batches to assess scalability and consistency. Results:Piloted against MRI reports, UODBLLM processed 1800 clinical documents with a 100% completion rate and an average processing time of 8.90 seconds per report. The token utilization averaged 2692 tokens per report, with an input-to-output ratio of approximately 13:2, resulting in a processing cost of US $0.009 per report. UODBLLM had consistent performance across 18 batches of 100 reports each and completed all processing in 4.45 hours. From each report, UODBLLM extracted 16 structured clinical elements, including prostate volume, prostate-specific antigen values, Prostate Imaging Reporting and Data System scores, clinical staging, and anatomical assessments. All extracted data were automatically validated against predefined schemas and stored in standardized JSON format. Conclusions:We demonstrated the successful integration of an LLM-based extraction system within an existing clinical outcomes database, achieving rapid, comprehensive data extraction at minimal cost. UODBLLM provides a scalable, efficient solution for automating clinical data extraction while maintaining protected health information security. This approach could significantly accelerate research timelines and expand feasible clinical studies, particularly for large-scale database projects.

BackgroundIdentifying metabolites associated with prostate cancer (Pca) aggressiveness may elucidate mechanisms underlying disease severity. Doing so in plasma and formalin-fixed, paraffin-embedded (FFPE) tissue may accelerate discovery. In this cross-sectional pilot study, we generated hypotheses for further exploration by assessing the association between plasma metabolites and the Gleason score of individuals with Pca and by evaluating the correlation between plasma and FFPE metabolite levels.MethodsWe examined plasma and FFPE samples from 10 individuals with Gleason score 7 (6 3 + 4, 4 4 + 3) and 9 individuals with Gleason score 9 (6 4 + 5, 3 5 + 4) tumors from a convenience sample of 19 men with Pca. We measured the relative abundance of polar metabolites at the time of radical prostatectomy. We used linear models of log2 fold changes to examine plasma metabolite levels relative to pathological tumor grade. Relationships among metabolite levels measured in plasma and FFPE tumor tissue within individuals across metabolites were examined using Pearson correlations.ResultsAmong the 18 selected plasma metabolites, which were chosen a priori because of their prior association with Pca aggressiveness, serine (p = 0.0051) and ornithine (p = 0.036) levels were higher in individuals with Gleason 9 than Gleason 7 Pca. After multiple testing correction, however, no associations were statistically significant. The median correlation between metabolite levels in plasma and FFPE tumor tissue was 0.45 (range: 0.40–0.53) for the 94 metabolites measured in both biospecimens.ConclusionsPlasma serine and ornithine levels demonstrated the largest differences between individuals with Gleason 7 and Gleason 9 Pca. Metabolite levels in FFPE prostate tissue samples were moderately correlated with plasma levels. Future studies with larger sample sizes are needed to further explore the hypotheses generated by this study.

BACKGROUND. Loss of the Y chromosome (LOY) is a frequent event in male tumors and has been linked to cancer progression. However, the degree of mosaic LOY (mLOY) within normal tissues from men with or without cancer remains uncharacterized. METHODS. Here we used a FISH-based assay targeting X- and Y-chromosome centromeres to perform a pan-organ analysis of mLOY in 1,000 male tissue samples from 405 individuals representing 11 organs. Automated image processing generated a quantitative FISH-based mLOY score (YchrFISH) that we validated against a transcriptomic surrogate of Y-chromosome dosage from RNA-seq data. RESULTS. mLOY burden varied by tumor type, with highest degree in colorectal carcinoma. Across tissue groups, YchrFISH scores declined progressively from normal tissues of cancer-free men to histologically normal tissues adjacent to cancer and carcinoma ( P < 0.0001). Paired analyses confirmed consistently greater mLOY in malignant compared with tumor-adjacent histologically normal tissue in different organs. Spatially resolved RNA-seq maps of bladders removed for cancer demonstrated a transcriptional gradient of Y-chromosome loss from normal urothelium through intraepithelial neoplasia to invasive carcinoma. CONCLUSION. mLOY gradients exist across histologically normal and malignant tissues, consistent with the concept of field cancerization. Our findings support epithelial mLOY as a biomarker of early malignant transformation and, to our knowledge, a previously unrecognized hallmark of male oncogenesis. FUNDING. NIH grants R35CA294022, P01CA163227, and P50CA97186 (the Pacific Northwest Prostate Cancer SPORE) and the Institute for Prostate Cancer Research.

BACKGROUND:A multimodal AI (MMAI) model has been validated in prostate biopsy specimens to guide treatment intensification in men receiving radiation. The MMAI has been explored to an extent for prostatectomy patients and has not yet been examined in relation to established genomic scores. METHODS:We applied the MMAI biopsy model to a tissue microarray (TMA) of 424 prostatectomy cases with long-term follow-up. MMAI scores were derived from digitized pathology images and clinical variables. Associations with biochemical recurrence and metastasis were tested using logistic and Cox regression, adjusting for the Cancer of the Prostate Risk Assessment (CAPRA) and genomic cell cycle progression (CCP) scores. RESULTS:MMAI scores were generated from one TMA spot for each of 414 patients (98%). At 10 years, recurrence-free and metastasis-free survival were 74% and 96%, respectively. In univariable models, MMAI was significantly associated with BCR (HR 1.04, 95% CI 1.02 to 1.06) and metastasis (HR 1.05, 95% CI 1.02 to 1.07). MMAI was not independently prognostic after adjustment for CAPRA, but remained significant when adjusted for CCP. Correlation between MMAI and either CAPRA or CCP was modest (r < 0.35). The model combining MMAI and CCP achieved the highest discrimination for metastasis (c-index 0.76), comparable to CAPRA (c-index 0.75). CONCLUSIONS:The MMAI score, originally developed for whole slide biopsy specimens, was prognostic when applied to a TMA of prostatectomy specimens despite not being designed for TMAs. Although not outperforming established clinical tools, it provided complementary information when combined with genomic data. The MMAI platform merits further refinement and validation in diverse clinical settings.

Genomic testing has potential to improve clinical decision-making for men with prostate cancer, but is understudied for active surveillance (AS), the standard management option for favorable-risk disease. We investigated whether Decipher scores are associated with AS outcomes in a cohort of patients on AS with at least two biopsies and a Decipher test. Decipher high-risk was defined as a score ≥0.6. Primary outcomes were any upgrading (any increase in grade group [GG]), major upgrading (GG ≥3), and unfavorable histology on subsequent biopsy. Multivariable Cox proportional-hazards regression models were generated for the cohort of 486 patients. On diagnostic biopsy, 78% had low risk and 22% had intermediate risk according to Cancer of the Prostate Risk Assessment (CAPRA) scores, and 12% had high risk according to Decipher. Decipher scores were associated with major upgrading after adjusting for CAPRA scores (hazard ratio [HR] 3.37, 95% confidence interval [CI] 1.17-9.69); high Decipher risk was associated with major upgrading after adjustment for either the CAPRA score (HR 2.00, 95% CI 1.14-3.49) or clinicodemographic variables (HR 2.65, 95% CI 1.36-5.17). The Decipher score was associated with unfavorable histology after adjustment for the CAPRA score (HR 3.68, 95% CI 1.03-13.08); high Decipher risk was associated with unfavorable histology after adjustment for clinicodemographic variables (HR 4.53, 95% CI 2.03-10.15) but not after adjustment for the CAPRA score. No association was observed for any upgrading. Deintensification of surveillance may be warranted for patients with lower Decipher risk.

323 Background: Disparities in prostate cancer (PCa) outcomes have been documented extensively across different racial and ethnic groups and various social environments, yet interventions to address these concerns remain elusive. We aim to understand how race and social vulnerability influence the likelihood of definitive treatment for men diagnosed with non-metastatic PCa treated in community urology practices. Methods: Men diagnosed with clinically localized PCa in the CaPSURE registry were geocoded, deidentified, and combined with Social Vulnerability Index (SVI) data, a census-tract measure of communities' disadvantage, based on participants’ home address at enrollment. Patients were categorized by race [Black or non-Black (White, Hispanic, Asian, Mixed, Other, Unknown)] and SVI status (high/low): Black-high SVI, Black-low SVI, non-Black-high SVI, and non-Black-low SVI. The outcome was primary treatment: definitive (RP, RT) vs non-definitive (AS/WW). Multinominal logistic regression analysis assessed the association between SVI, race, and odds of definitive treatment, adjusted for age at diagnosis, education, income, insurance, comorbidities, BMI, smoking, alcohol use, year of diagnosis, clinical site, job type, and US Census subregion. Models were run separately for low and intermediate/high clinical PCa. A p-value <0.05 was statistically significant. Results: In total, 7854 men were identified with 714 (9%) Black and 3573 (45%) residing in high SVI communities; 72% of Black men had high SVI compared to 43% of Non-Black men (p<0.01). High risk disease was more common for Black men with high SVI compared to others (20% vs 9% for Black-low SVI vs 12% Non-Black high SVI vs 11% Non-Black low SVI, p<0.01). For low-risk disease, Black men with high SVI had lowest odds of definitive treatment compared to non-Black men with low SVI (OR 0.54, 95% CI 0.32-0.90); odds for other groups did not differ significantly. For intermediate/high risk disease, Black men with low SVI had the lowest odds of definitive treatment compared to non-Black men with low SVI (OR 0.24, 95% CI 0.10-0.56); odds for the other groups did not differ significantly. Conclusions: This study highlights the significant impact of social vulnerability on prostate cancer treatment outcomes among Black men, particularly those with high SVI, despite access to community urologic care. Black men with high SVI experiencing lower odds compared to their non-Black counterparts, even after adjustment for clinical risk, insurance coverage, comorbidities, geographic disparities, and age at diagnosis also contribute to treatment outcomes, emphasizing the intricate interplay of these factors. The unique combination of CaPSURE and SVI allows for the comprehensive assessment and geographic localization of multilevel drivers of treatment disparities which persist even in men receiving urologic care.

43 Background: Prostate-specific membrane antigen (PSMA) PET imaging is recommended for patients with biochemical recurrence (BCR) after radical prostatectomy (RP). This study describes the early clinical impact of PSMA-PET compared to conventional imaging (CT, MRI, or bone scintigraphy) in a contemporaneous period. Methods: This retrospective study includes patients with BCR after RP (defined as PSA≥0.2 after undetectable PSA post-operatively) who underwent PSMA-PET or conventional imaging (CI) between 2010-2023. The primary outcome was positive detection rate. The secondary outcomes were rates of salvage treatment (XRT±ADT) and second BCR. Cox proportional hazards modeling was used to predict risk of salvage treatment and second BCR. Results: 217 post-RP patients were included. Of the 146 patients who underwent PSMA-PET and the 71 who underwent CI, 77 (52.7%) and 7 (9.9%) patients had positive imaging findings at a median PSA value of 0.23 (IQR 0.21-0.33) and 0.25 (IQR 0.25-0.44), respectively (p=0.36). 48 patients received both imaging modalities; 4 had positive results in both PSMA-PET and CI, while 26 were positive for PSMA-PET only, and 2 were positive on CI only. Patients underwent salvage therapy at lower median PSA values in the PSMA-PET group (0.40 vs 0.51, p<0.01). On unadjusted survival analysis, there were no differences in salvage therapy rates or second BCR rates between PSMA-PET or CI groups. On multivariable Cox proportional hazards modeling, year of PSA recurrence (HR 0.89, 95% CI 0.84-0.95), BMI (HR 1.45, 95% CI 1.02-2.07), ≥GG3 (HR 1.77 95% CI 1.3-2.4), and undetectable PSA≥6 months after RP (HR 0.50 95% CI 0.34-0.72) were associated with salvage treatment, but imaging modality was not. High genomic risk scores (Decipher>0.60) may be associated with second BCR (HR 2.03 95% CI 0.99-4.18, p=0.05) while other variables were not. Conclusions: PSMA-PET usage increased over the period of this study with higher sensitivity at lower median PSA at BCR. Imaging modality did not predict rates of salvage therapy or second BCR, while other clinical risk factors such as adverse pathology and high genomic risk score did. Analysis with a historical cohort is pending.

Introduction High-intensity focused ultrasound (HIFU) is a type of focal therapy that uses magnetic resonance imaging (MRI) and transrectal ultrasound (TRUS) imaging to specifically target areas affected by prostate cancer (PCa), with careful attention to avoid normal tissues. HIFU has been shown in prior studies to have moderate PCa control outcomes with little impact on sexual or urinary side effects. A significant concern limiting widespread adoption of HIFU is inadequate treatment of disease and recurrence. Here we share institutional biopsy-proven recurrence outcomes following HIFU, and identify clinicodemographic factors associated with in-field and overall recurrence. Methods All men who underwent robotic HIFU (FocalOne) at the University of California, San Francisco and underwent 1-year post-procedure MRI-fusion biopsy were included. The primary outcomes were in-field recurrence, defined as biopsy-diagnosed PCa greater than or equal to Gleason Grade 2 (≥GG2) in the same sextant of HIFU treatment, and overall recurrence, defined as ≥GG2 diagnosed anywhere in the prostate. Secondary outcomes were urinary and sexual quality of life (QOL) changes, measured by the International Prostate Symptom Score (IPSS) and Sexual Health Inventory for Men (SHIM) score. Multivariable logistic regression models were used to estimate associations between patient characteristics and risk of recurrence, controlling for age, Gleason grade, number of positive biopsy cores, MRI Prostate Imaging Reporting and Data System (PI-RADS) score, genomic risk score (based on Decipher, Oncotype Dx, and Prolaris scores), and PSA density. Results There were 102 men included in analyses who had 1-year post-HIFU biopsy follow-up. In-field and overall recurrence on 1-year biopsy was 42% and 56%, respectively. IPSS and SHIM scores were not significantly different between pre-HIFU and one-year post-HIFU (p>0.05). GG3 or greater disease was borderline associated with greater risk of in-field recurrence (OR=2.84; 95% CI 1.00-8.07; p=0.05; Table 1), and pre-HIFU PSA was associated with greater risk of overall recurrence (OR=1.19; 95% CI 1.04-1.37 p=0.01; Table 1). Conclusions In-field and overall recurrence rate on biopsy one year after HIFU was found to be 42% and 56%, respectively. GG3 or greater disease trended towards in-field disease recurrence and increased pre-HIFU PSA was associated with overall disease recurrence. No significant changes in sexual or urinary symptoms were observed. These findings emphasize the importance of careful patient selection for HIFU, which has potential for cancer control with minimal side effects in the appropriate PCa patient.