PD52-06 COMPUTATIONAL IMMUNOGENOMIC MODELING OF IMMUNOEVASIVE AGGRESSIVE PROSTATE CANCER

Abstract

You have accessJournal of UrologyProstate Cancer: Markers II (PD52)1 Apr 2020PD52-06 COMPUTATIONAL IMMUNOGENOMIC MODELING OF IMMUNOEVASIVE AGGRESSIVE PROSTATE CANCER Howard Korman*, Kirk Wojno, Amin Kassis, Ricardo Henao, Jason Hafron, Kenneth Kernen, Geoffrey Erickson, Harry Stylli, Shrikant Mane, Angela Busta, Elyse Marriott, Rima Tinawi-Aljundi, Mathew Putzi, and Philip Kantoff Howard Korman*Howard Korman* More articles by this author , Kirk WojnoKirk Wojno More articles by this author , Amin KassisAmin Kassis More articles by this author , Ricardo HenaoRicardo Henao More articles by this author , Jason HafronJason Hafron More articles by this author , Kenneth KernenKenneth Kernen More articles by this author , Geoffrey EricksonGeoffrey Erickson More articles by this author , Harry StylliHarry Stylli More articles by this author , Shrikant ManeShrikant Mane More articles by this author , Angela BustaAngela Busta More articles by this author , Elyse MarriottElyse Marriott More articles by this author , Rima Tinawi-AljundiRima Tinawi-Aljundi More articles by this author , Mathew PutziMathew Putzi More articles by this author , and Philip KantoffPhilip Kantoff More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000000954.06AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Peripheral blood CD14/CD2 RNA expression ratios have shown associations with prostate biopsy features. The objectives of this study are: 1) To develop a computational immunogenomic model of immunoevasive aggressive prostate cancer and validate this model on an independent data set, 2) determine if dimensionality reduction strategies are effective in minimizing model overfitting, 3) explore pathway associations and determine links with biopsy features, and 4) compare this optimized model with mainstream published models predicting biopsy features. METHODS: Peripheral blood samples from 696 urology patients between 40 and 75 years of age were obtained under IRB approval, prior to their scheduled prostate biopsy. Whole transcriptome RNA sequencing was performed on two purified cell types (CD14 and CD2) and RNA-sequencing expression levels were used to develop predictive models correlating to biopsy results using 50-fold cross validation. An independent validation set of 301 patients was used to confirm the results. Receiver Operating Characteristic (ROC) curves were created to compute p-values comparing the area under the curve (AUC) of individual predictive variables to the overall model, as well as to published mainstream models. RESULTS: In the discovery set the AUC of the ROC was 0.75 for positive versus negative biopsy results (BR), 0.82 for Gleason grade group (GG), 0.84 for number of positive biopsy needle cores (CP) of, 0.81 for maximum involvement of a core by tumor (MI), and 0.80 for the aggregate of the biopsy features (ABF). The degree of overfitting was estimated by the drop of performance seen in the independent validation set and was not statistically significant for any of the output variables. Our Model performs significantly better than mainstream models such as the PCPT and the PBCG risk calculators. Interesting pathway associations between tumor volume estimates and antigen processing, along with antigen peptide processing by MHC class II, are observed. CONCLUSIONS: A whole transcriptome RNA-seq algorithm based on clinical immunogenomic parameters determined by the analysis of peripheral blood samples from pre-biopsy prostate cancer patients predicts prostate biopsy features significantly better than established and widely used algorithms. The performance characteristics of this model have been validated on an independent data set with no significant overfitting or drop in performance noted. This study demonstrates the power of RNA expression patterns of pure populations of peripheral blood immune cells to predict prostate biopsy outcome. Source of Funding: OncoCell MDx © 2020 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 203Issue Supplement 4April 2020Page: e1091-e1092 Advertisement Copyright & Permissions© 2020 by American Urological Association Education and Research, Inc.MetricsAuthor Information Howard Korman* More articles by this author Kirk Wojno More articles by this author Amin Kassis More articles by this author Ricardo Henao More articles by this author Jason Hafron More articles by this author Kenneth Kernen More articles by this author Geoffrey Erickson More articles by this author Harry Stylli More articles by this author Shrikant Mane More articles by this author Angela Busta More articles by this author Elyse Marriott More articles by this author Rima Tinawi-Aljundi More articles by this author Mathew Putzi More articles by this author Philip Kantoff More articles by this author Expand All Advertisement PDF downloadLoading ...