360 - Genomic Classifier and Multimodal AI Biomarker in Localized Prostate Cancer: Two Sides of the Same Coin
Presenter(s)
P. Trivedi1, S. Awasthi2, R. Putney1, E. Katende1, A. Serna3, R. Smith4, A. Hakansson5, Y. Liu6, E. Davicioni7, Y. Ren8, S. Tang8, D. Croucher9, J. Dhillon1, J. Park1, R. Pessoa3, M. Poch10, D. Grass11, J. Pow-Sang1, and K. Yamoah11; 1H. Lee Moffitt Cancer Center & Research Institute, Tampa, FL, 2H. Lee Moffitt Cancer Center and Research Institute, Department of Cancer Epidemiology, Tampa, FL, 3H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, 4Moffitt Cancer Center, Tampa, FL, 5Veracyte Inc.,, South San Francisco, CA, 6Veracyte, San Diego, CA, 7Veracyte Inc., San Diego, CA, 8Artera, Inc., Los Altos, CA, 9ArteraAI, Los Altos, CA, 10Department of Genitourinary Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, 11Department of Radiation Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL
Purpose/Objective(s): The predictive accuracy of prostate cancer (PCa) specific outcomes using conventional clinical risk classifiers remain suboptimal. The emergence of personalized biomarkers such as the Decipher prostate genomic classifier (GC) and Artera multimodal AI (MMAI) have shown superior prognostic performance by more accurately identifying aggressive subsets of PCa. Here, we report on the clinical and biological correlation between GC and MMAI as predictors of metastasis among patients with localized PCa.
Materials/Methods: We performed a retrospective analysis of patients with localized PCa who had available genomic data and digital histopathologic images from radical prostatectomy (RP) specimens. GC scores were generated using RNA sequencing data whereas MMAI scores were generated utilizing histopathological image features combined with clinicopathologic parameters. Both GC and MMAI scores were evaluated as continuous variables. The primary aim was to establish the correlation between GC and MMAI biomarkers for predicting biochemical recurrence (BCR), and metastasis using linear regression. Next, intergene correlational matrix was performed using derived gene signatures to identify distinct biological pathways that associate with GC and MMAI scores in predicting metastasis.
Results: Final analytical cohort included 190 evaluable cases with both GC and MMAI information, including 40 cases with BCR and 8 metastatic events with a median follow up of 5.8 years. The overall median GC and MMAI scores were 0.34 and 0.30, respectively. Linear regression demonstrated a weakly positive association between the two scores (R2 = 0.19, 95%CI 0.46-0.84) in the overall cohort. Among patients with BCR, the association remained weak (R2 = 0.36, 95%CI 0.37-0.95), and no correlation (R2 = 0.0001, 95%CI 0.00-0.87) was observed among patients with metastasis. In correlative analysis, GC and MMAI scores showed no strong associations overall, as well as among patients with BCR. Intergene correlational matrix performed within the metastasis cohort revealed a strongly positive correlation between MMAI score and immune-related pathways (immune estimate: r=0.93, IFN-a: r=0.78, IFN-?: r=0.86, immune T-cells: r=0.81), whereas GC showed strongly positive correlation with angiogenic signaling (angiogenesis: r=0.88, endothelial cells: r=0.93) and immune-related (CD4: r=0.78, CD8: r=0.79) pathways. Importantly, there was minimal overlap between the specific target genes that constitutes the distinct signature pathways.
Conclusion: Our study demonstrates that Decipher GC and Artera MMAI scores are weakly correlated. However, both scores are strongly, yet independently associated with distinct biological pathways in patients who developed metastasis. Further studies are required to determine whether the integration of these biomarkers may better define the biological spectrum of aggressive PCa.