Lack of efficient noninvasive biomarkers for differentiating prostate cancer (PCa) and benign prostate hyperplasia (BPH) is a serious concern for men's health worldwide. In this study, we aimed to improve the diagnostic capability of the existing noninvasive biomarkers for PCa. GC-MS-based untargeted metabolomics was employed to analyze plasma samples for 41 PCa patients and 38 BPH controls. Both univariate and multivariate statistical analyses were performed to screen for differential metabolites between PCa and BPH, followed by the selection of potential biomarkers through machine learning. The chosen candidate biomarkers were then verified by targeted analysis and transcriptome data. The results showed that twelve metabolites were significantly dysregulated between PCa and BPH, three metabolites including L-serine, myo-inositol, and decanoic acid could be potential biomarkers for discriminating PCa from BPH. Most importantly, ROC curve analysis demonstrated that the involvement of the three potential biomarkers has increased the area under the curve (AUC) value of cPSA and tPSA from 0.542 and 0.592 to 0.781, respectively. Therefore, it was concluded that the involvement of L-serine, myo-inositol, and decanoic acid can largely improve the diagnostic capability of the commonly used noninvasive biomarkers in the clinic for differentiating PCa from BPH.
Keywords: Benign prostate hyperplasia; Biomarker; Metabolomics; Prostate cancer; Transcriptome.
Copyright © 2021 Elsevier B.V. All rights reserved.