A blood-based multi-mRNA liquid biopsy with >90% accuracy for diagnosis and assessment of prostate cancers.

Abstract

5574 Background: There are a paucity of blood-based biomarkers with clinical utility for prostate cancer (PCa). We developed a circulating mRNA (27-gene) prostate cancer signature to diagnose and manage PCa. Methods: Gene identification: Publicly available PCa transcriptome sets ( n= 1,159 samples) were evaluated and compared with normal blood-based transcriptomes using gene co-expression network enrichment, differential expression and functional enrichment analyses to identify candidate markers. Gene expression evaluation: Seven PCA cell lines and two normal prostate epithelial lines were used to assess candidate genes. Marker genes were determined in PCa tumor tissue ( n= 50) and validated in the TCGA-PRAD ( n= 500) dataset. Blood gene expression: Set #I: PCA: n= 132, BPH: n= 44, controls n= 55. Set #II: n= 50 (biochemical recurrence [BCR]). We constructed an artificial intelligence PCa model using classification algorithm analyses. Scoring: normalized algorithmically analyzed gene expression (0 to 100), positive score >20. PSA: BPH ( n= 44) and PCa ( n= 132). Clinical score assessment: Surgical cohort: ( n= 47), samples: pre-surgical and post: 1 week - 14 months. Statistics: Kruskal-Wallis, Pearson-correlation, Fisher’s and AUROC analyses (Mean±SEM). Results: Transcriptomic analysis identified 27 candidates. Cell lines/tissue: Expression levels were significantly elevated ( p< 0.001, 2.1-35.8-fold) in cell lines and PCa surgical samples. All 27 markers were confirmed in TCGA-PRAD samples (average TPM: 58 to 10,366). Blood: In Set#I, levels in PCa were 47±2 ( p< 0.0001) compared to BPH (19±1) and controls (18±0.5); AUROC: 0.92 (BPH) and 0.94 (controls), with an accuracy of 85-88%, a sensitivity of 86% and specificities 82 and 93%. For PSA, the AUROC (PCa vs. BPH) was 0.51 ( p= 0.88). PSA was positive in 86% of BPH and was > 10ng/ml in 30%. PSA was positive in 83% of PCa and > 10ng/ml in 40% (Fisher’s p= 0.28). PSA accuracy ( > 10ng/ml) was 48%. Levels in Set#II (BCR) were 44±3. ProstaTest-was positive in 48 (96%). Surgical cohort ( n= 47): Prostatest accuracy 100% pre-surgery. Resection decreased levels (KW-statistic: 57.4, p< 0.0001) from 52±1 to 23.5±2. Conclusions: A 27-gene blood signature was developed for PCa that exhibited a diagnostic accuracy of 92%; significantly better than PSA (48%, p< 0.0001). Surgical resection significantly ( p< 0.0001) decreased levels. Biochemical recurrence was accurately detected (96%). A multi-gene prostate cancer liquid biopsy is likely to have clinical utility in both diagnosis and monitoring of PCa.