A multiomic approach for detecting clinically significant prostate cancer in seminal fluid: Results from a large multicenter clinical study.

Abstract

e17135 Background: While Prostate Specific Antigen (PSA) is the standard of care for prostate cancer screening, it is highly controversial due to the performance characteristics of the test. Magnetic Resonance Imaging (MRI) is becoming increasingly used as a reflex test for high PSA, but there are limitations with cost, disparities in access, variations in interpretation, and moderate sensitivities/specificities. Given that the prostate is a reproductive organ and a large portion of seminal emissions are generated from the prostate, this study aimed to investigate whether a novel prostate cancer screening platform using cell free nucleic acids from seminal plasma could improve the detection of clinically significant prostate cancer. Methods: Over a 4-month period, 276 patients scheduled to undergo a prostate biopsy were enrolled across 12 sites. Semen samples were collected at home by men aged 40 yrs and older and shipped to the Fellow Health laboratory. Seminal plasma was prepared by centrifugation and cell free nucleic acids extracted. DNA methylation libraries were prepared using the Twist methylome target enrichment panel and RNA libraries were prepared using Watchmaker kits. All libraries were sequenced on a NovaSeq X. Clinical data (including PI-RAD score, cancer status and Grade Group (GG)) were merged with the sequencing results and machine learning tools employed for modeling predictive algorithms. Results: Median age was 64 (IQR=10 years), median PSA was 6.1 (IQR=3.925ng/ml), 79% of subjects were white, 45% had BPH, 12.5% were vasectomy positive and 66% received MRI prior to biopsy. 180 valid semen samples were received prior to the cutoff date, of which 146 were evaluable for both RNA and DNA analysis. The well-known prostate tissue markers PSA and TMPRSS2 were highly abundant in seminal fluid with mean normalized transcripts per million of 1597 and 397.5 respectively. Using healthy controls to set the limit of blank, 45.1% of cancer patients harbored an ETS fusion (ERG, ETV1, ETV4, ETV5, FLI1). The data was split into a training set (n=115) and an independent test set (n=31). With a panel of 66 expression markers, 26 methylation markers and PSA, the AUC in the test set was 0.83 for differentiating clinically significant prostate cancer (GG2+) from no cancer (<=GG1). Conclusions: In this cohort, the use of MRI still resulted in unnecessary biopsies 69.4% of the time. Meanwhile, we demonstrate that our multiomic approach using cell free DNA and RNA achieves high accuracy for identifying clinically significant prostate cancer. At-home semen collection offers a convenient alternative to in-lab testing and may have the potential to augment MRI in post-PSA reflex testing.