You have accessJournal of UrologyCME1 Apr 2023MP40-03 CIRCULATING CHROMOSOME CONFORMATION SIGNATURES ENHANCE THE ACCURACY OF PSA FOR DETECTING PROSTATE CANCER - RESULTS FROM PERFORMANCE OF THE EPISWITCHTM PCA TEST IN THE PROSTAGRAM TRIAL Jiten Jaipuria, Dmitri Pchejetski, David Eldred-Evans, Martin J Connor, Hashim U Ahmed, Alexandre Akoulitchev, Ewan Hunter, and Mathias Winkler Jiten JaipuriaJiten Jaipuria More articles by this author , Dmitri PchejetskiDmitri Pchejetski More articles by this author , David Eldred-EvansDavid Eldred-Evans More articles by this author , Martin J ConnorMartin J Connor More articles by this author , Hashim U AhmedHashim U Ahmed More articles by this author , Alexandre AkoulitchevAlexandre Akoulitchev More articles by this author , Ewan HunterEwan Hunter More articles by this author , and Mathias WinklerMathias Winkler More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003278.03AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: PSA screening for prostate cancer (PCa) has a low overall accuracy despite high specificity. Novel blood biomarkers thus are prime candidates for improvement. 3-dimension chromatin conformations (CCCs) are known epigenetic regulators that manifest early in PCa tumorigenesis. EpiSwitch™ is a blood sample-based epigenetic assay developed by us which algorithmically analyses CCCs. We previously found diagnosed high-risk PCa associated with CCCSs in the loci of HSD3B2, VEGFC, APAF1, BMP6, ERG, MSR1, MUC1, ACAT1 and DAPK1 genes. Here, we evaluate the ability of EpiSwitch™ to detect PCa in a screening cohort of patients in the Prostagram trial. METHODS: We created a population cohort (n=147) of men, 50-69 years of age, comprising 109 volunteers participating in PCa screening in the Prostagram trial, and enriched it hospital tissue bank samples of 29 men with established localised PCa, and 9 cancer-negative controls. The cohort was then randomly split into training and testing groups in a 1:2 ratio. Samples were tested for PSA and the presence of CCCSs via the EpiSwitch™ assay. Batch adjustment by reference alignment procedure was used to control for batch effects. Feature engineering of the EpiSwitch™ markers with PSA was performed by recursive feature elimination using linear discriminant analysis, xgbtree, xgblinear, decision trees and random forest libraries in the R program. Shapley additive explanations values were used to quantify the contribution of each feature within the model. XGBoost algorithm model was used for final test optimisation and ultimately the EpiSwitch™PCa test was created. We compared combinations of - binary PSA threshold (>3 ng/ml), continuous PSA, and EpiSwitch™PCa-in terms of accuracy, positive (PPV) and negative predictive values (NPV) to determine PCa in the Prostagram cohort. RESULTS: ERG21 marker was discarded during the test optimisation to create the EpiSwitch™PCa assay. PSA>3ng/ml had a low PPV of 0.14 and a high NPV of 0.93 in the Prostagram cohort. Episwitch™PCa alone showed a PPV of 0.91 and NPV of 0.32. EpiSwitch™PCa combined with PSA>3ng/ml had a PPV of 0.81 and NPV of 0.78. However, EpiSwitch™PCa combined with continuous PSA yielded a remarkable PPV of 0.92, and NPV of 0.94, with an overall accuracy of 0.94. CONCLUSIONS: Combining the PSA test with specific CCCs in the blood has a potential for rapid PCa diagnosis with very high accuracy and low cost justifying further research in a prospective blinded cohort study. Source of Funding: Cost of the epigenetic assay was funded by Oxford biodynamics. Prostagram study was funded by the Wellcome trust. Data was analysed by Oxford biodynamics within the a priori statistical analysis plan of the Prostagram study. Analysis was performed with oversight of an independent academic statistician from the Prostagram study trial management group. © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e545 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information Jiten Jaipuria More articles by this author Dmitri Pchejetski More articles by this author David Eldred-Evans More articles by this author Martin J Connor More articles by this author Hashim U Ahmed More articles by this author Alexandre Akoulitchev More articles by this author Ewan Hunter More articles by this author Mathias Winkler More articles by this author Expand All Advertisement PDF downloadLoading ...
Read full abstract