Reliable liquid biopsy-based tools able to accurately discriminate prostate cancer (PCa) from benign prostatic hyperplasia (BPH), when PSA is within the “gray zone” (PSA 4–10), are still urgent. We analyzed plasma samples from a cohort of 102 consecutively recruited patients with PSA levels between 4 and 16 ng/ml, using the SANIST-Cloud Ion Mobility Metabolomic Mass Spectrometry platform, combined with the analysis of a panel of circulating microRNAs (miR). By coupling CIMS ion mobility technology with SANIST, we were able to reveal three new structures among the most differentially expressed metabolites in PCa vs. BPH. In particular, two were classified as polyunsaturated ceramide ester-like and one as polysaturated glycerol ester-like. Penalized logistic regression was applied to build a model to predict PCa, using six circulating miR, seven circulating metabolites, and demographic/clinical variables, as covariates. Four circulating metabolites, miR-5100, and age were selected by the model, and the corresponding prediction score gave an AUC of 0.76 (C.I. = 0.66–0.85). At a specified cut-off, no high-risk tumor was misclassified, and 22 out of 53 BPH were correctly identified, reducing by 40% the false positives of PSA. We developed and applied a novel, minimally invasive, liquid biopsy-based powerful tool to characterize novel metabolites and identified new potential non-invasive biomarkers to better predict PCa, when PSA is uninformative as a tool for precision medicine in genitourinary cancers.