3397 - Cross-Continental Transcriptomic Evaluation of a Multimodal AI-Based Biomarker in Localized Prostate Cancer
Presenter(s)
S. Webking1, P. Trivedi2, R. Putney2, E. Katende2, A. Serna3, R. C. Smith4, A. Abrahams5, Y. Ren6, D. Mukherjee7, S. Tang6, D. Croucher7, E. Stewart8, A. Esteva7, A. Rishi9, R. Pessoa10, M. Poch10, J. Mensah5, J. Dhillon11, J. Park2, M. E. Tharp II12, J. Yarney13, G. D. D. Grass14, J. Pow-Sang10, and K. Yamoah14; 1Moffitt Cancer Center, Tampa, FL, United States, 2H. Lee Moffitt Cancer Center & Research Institute, Tampa, FL, 3Moffitt, Tampa, FL, 4Moffitt Cancer Center, Tampa, FL, 5Korle Bu Teaching Hospital, Accra, Ghana, 6Artera, Inc., Los Altos, CA, 7ArteraAI, Los Altos, CA, 8Artera, Los Altos, CA, 9H. Lee Moffitt Cancer Center and Research Institute, Department of Radiation Oncology, Tampa, FL, 10Department of Genitourinary Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, 11Department of Anatomic Pathology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, 12Indiana University, Indianapolis, IN, 13National Centre for Radiotherapy and Nuclear Medicine, Accra, Ghana, 14Department of Radiation Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
Purpose/Objective(s):
Prostate cancer (PCa) has the largest heterogeneity in outcomes among different populations globally. Even after adjusting for social determinants of health, PCa incidence and mortality rates remain divergent among different groups. Therefore, there is a pressing need to identify biomarkers that demonstrate biological robustness across heterogeneous populations. The emergence of personalized biomarkers such as the Artera multimodal artificial intelligence (MMAI) has shown superior ability to aid in the detection and prognostication of aggressive PCa. Here, we report the pathway- and gene- level evaluation of the MMAI biomarker across multiple patient cohorts, using the HTG EdgeSeq platform.Materials/Methods: We retrospectively analyzed two independent transcriptomic datasets: native African (NAM) patients from Africa and African American (AAM) plus European American (EAM) patients from a U.S. tertiary cancer center. Samples were processed on HTG EdgeSeq using the oncology biomarker panel and matched by Gleason score. Biopsy MMAI scores were generated by applying deep learning to digitized H&E slides, integrating histopathologic features with clinical variables. Gene set enrichment analysis (GSEA) identified pathways enriched in MMAI high-risk patients and assessed cross-cohort concordance. Differential expression analyses identified genes consistently associated with MMAI risk-status across groups.
Results:
The final cohort consisted of 90 patients comprised of 32 NAM, 18 AAM, and 40 EAM. GSEA revealed significant enrichment of oncogenic and proliferative pathways, including E2F targets, Epithelial-mesenchymal transition, and KRAS signaling Up, in MMAI high-risk patients among both cohorts. Among these overlapping pathways, we examined gene-level directionality across cohorts. Aminopeptidase N (ANPEP) was down-regulated, while SFRP4 was up-regulated in high-risk MMAI group across cohorts suggesting aggressive tumor biology. Differential expression analysis further identified additional genes significantly associated with MMAI status in both cohorts. Several genes demonstrated consistent directionality across all three groups, aligning with established literature linking these genes to aggressive PCa phenotypes. CD38 and HMGCS2, which have been shown to be down-regulated in aggressive PCa, were significantly down-regulated in MMAI high-risk samples, while SMC3, TLR4, NCAPD3, and NFKBIZ, all associated with PCa aggressiveness, were found to be significantly up-regulated in the MMAI high-risk group. Overall, gene expression patterns tracked with MMAI risk categorization in the different populations.Conclusion:
Our study findings demonstrate the biological robustness and cross-population generalizability of the MMAI, supporting its utility for risk stratification among diverse PCa populations.