A Serum Metabolomic In Vitro Diagnostic Multivariate Index Assay for Prostate Cancer Detection

XIAOLING ZANG; CHRISTINA M. JONES; TRAN Q. LONG; MARÍA EUGENIA MONGE; MANSHUI ZHOU; L. DEETTE WALKER; ROMAN MEZENCEV; ALEXANDER GRAY; JOHN F. MCDONALD; FACUNDO M. FERNÁNDEZ

Baltimore

Conferencia; 62nd ASMS Conference on Mass Spectrometry and Allied Topics; 2014

American Society for Mass Spectrometry

Introduction Prostate cancer (PCa) is the second leading cause of cancer-related mortality in men worldwide. The prevalent PCa diagnosis method using Prostate Specific Antigen (PSA) screening suffers from low specificity and over-diagnosis. These disadvantages have led to increased interest of using untargeted metabolite profiling to discover new differential metabolic biomarkers that could improve PCa diagnosis performance. However, global untargeted metabolic profiling of PCa patients is still in an early stage, and there is no biomarker panel currently used for clinical testing. In this study, ultra high performance liquid chromatography coupled to high resolution tandem mass spectrometry (UHPLC-MS/MS) and machine learning methods were used to profile and identify metabolites in blood sera which discriminate PCa patients from healthy individuals. Methods Age-matched blood serum samples from 64 PCa patients and 50 healthy individuals were profiled by UHPLC-MS, using a Waters ACQUITY Ultra Performance LC system, fitted with a Waters ACQUITY UPLC BEH C18 column and coupled to a Synapt G2 High Definition Mass Spectrometry (HDMS) system. UHPLC-MS/MS experiments were performed with applied voltages between 5 and 50V, using argon as collision gas. Metabolites were extracted using a mixture of acetone, acetonitrile and methanol (1:1:1 v/v). Mass spectra were acquired in negative ionization mode in the m/z range of 50-1750. Metabolic features were extracted using MassLynx v4.1. Support vector machines were implemented to develop an in vitro diagnostic multivariate index assay (IVDMIA) for sample classification. Preliminary Data A total of 480 features (Rt, m/z pairs) were quantitatively detected in serum metabolome profiles from the entire cohort. The developed IVDMIA successfully predicted the presence of PCa with 93.0% accuracy, 94.3% specificity and 92.1% sensitivity using only a panel of 40 metabolic spectral features. Our IVDMIA proved to have a better detection performance than the prevalent PSA test for PCa patients in the cohort, as 33% of the PCa patients had PSA values lower than the commonly used cutoff point of 4.0 ng mL-1. These results highlight that a combination of multiple discriminant features yields higher predictive power for PCa diagnosis than univariate analysis of a single protein marker. Within the discriminant panel, 31 metabolites were identified by MS and MS/MS, with 10 further confirmed chromatographically by standards. Numerous identified metabolites were mapped to metabolic pathways of fatty acids, amino acids, lysophospholipids, and bile acids. The most relevant finding is that many differential metabolites belong to the steroid hormone biosynthesis pathway, which supplies androgens such as testosterone and 5α-dihydrotestosterone to support the growth of androgen-dependent PCa. These results provide further insights into the metabolic alterations associated with the disease. If higher throughput analysis in addition to lower analysis cost and complexity are desired, 13 metabolites that were found to be present in 90% of the entire sample cohort would still provide high classification sensitivity (88.3%), specificity (80.3%), and accuracy (85.0%) for PCa patients and healthy individuals. These results show great promise towards clinical implementations. Novel Aspect High classification accuracy suggests this approach is a promising tool for accurate PCa diagnosis.