Abstract INTRODUCTION: Previous studies among largely European American (EA) populations have identified prostate tumor gene expression signatures that are associated with a tumor’s Gleason score, the morphologic grade of prostate cancer (PCa) that is strongly associated with PCa mortality. Here, we aimed to assess the ability of two published gene signatures (Sinnott et al, 2017 and Jhun et al, 2017) to distinguish between high- and low-grade tumors among African American (AA) men with PCa, who bear a disproportionate burden of PCa mortality. Among the genes that were previously associated with Gleason score or poor PCa outcomes (including components of the two signatures), we also aimed to identify the most important genes for predicting high- vs. low-grade tumors among AA men and to develop a new signature based on these genes. METHODS: We accessed formalin-fixed, paraffin-embedded prostate tumor tissue for 189 AA men with PCa who underwent radical prostatectomy at the University of Maryland Medical Center from 1992-2021. Tumor grade was determined from pathology reports. We generated tumor RNA expression data using the ThermoFisher Human Clariom D array. We assessed the ability of the published gene signatures to distinguish Gleason Grade Groups 4 or 5 (GG4/5, high-grade) from Gleason Grade Group 1 (GG1, low-grade) in our study population by computing the Area Under the Curve (AUC) and 95% confidence interval (95% CI). We also separately modeled individual genes included in the signatures and other genes previously associated with GG or poor PCa outcomes (total of 84 genes) in relation to GG using ordinal logistic regression. Models were adjusted for age at surgery and year of surgery. We adjusted for multiple comparisons using the Benjamini-Hochberg False Discovery Rate (FDR) method. RESULTS: Twenty-eight percent of participants were classified as GG1, 49% as GG2, 13% as GG3, and 10% as GG4/5. The AUC for predicting GG4/5 vs. GG1 in our study population was 0.53 (95% CI: 0.38-0.67) for the first signature and 0.65 (95% CI: 0.50-0.79) for the second compared to the published AUC of 0.76. Of the 84 genes previously associated with GG or poor PCa outcomes, 16 genes remained significantly associated with GG in the ordinal logistic regression models after multiple comparison adjustment (FDR-adjusted p-values0.05). The most significant gene was A-kinase anchoring protein 19 (AKAP19; p-value=2.7x10-5; FDR-adjusted p-value=0.002). A new signature based on these 16 genes using elastic net showed relatively high predictive ability for GG4/5 vs. GG1 (AUC: 0.85, 95% CI: 0.76-0.94). CONCLUSION: Gene signatures identified among predominantly EA men with PCa had modest ability to distinguish between high- and low-grade tumors among AA men in our study, but a subset of the genes had high predictive ability in this population. Future directions will include assessing the ability of this new signature to predict PCa mortality among AA men. Results highlight the importance of including diverse populations in research to identify prognostic signatures for PCa. Citation Format: Ebuka Onyenobi, Jessica Yau, Yuji Zhang, Teklu B. Legesse, Gary Rose, Guangjing Zhu, Allen Burke, Ashley Johnson, Kimberly Clark, Nicholas Ambulos, Jing Yin, Jong Y. Park, Soren M. Bentzen, Arif Hussain, Joanne Dorgan, Lorelei A. Mucci, Kathryn H. Barry. Predictive ability of prostate tumor gene expression signatures for high- and low-grade tumors among African American men abstract. In: Proceedings of the 18th AACR Conference on the Science of Cancer Health Disparities; 2025 Sep 18-21; Baltimore, MD. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2025;34(9 Suppl):Abstract nr A081.
Onyenobi et al. (Thu,) studied this question.