This work focuses on the influence of the selected LC-HRMS platform on the final annotated compounds in non-targeted metabolomics. Two platforms that differed in columns, mobile phases, gradients, chromatographs, mass spectrometers (Orbitrap [Platform#1] and Q-TOF [Platform#2]), data processing and marker selection protocols were compared. A total of 42 wines samples from three different protected denomination of origin (PDO) were analyzed. At the feature level, good (O)PLS-DA models were obtained for both platforms (Q(2)[Platform#1]=0.89, 0.83 and 0.72; Q(2)[Platform#2]=0.86, 0.86 and 0.77 for Penedes, Ribera del Duero and Rioja wines respectively) with 100% correctly classified samples in all cases. At the annotated metabolite level, platforms proposed 9 and 8 annotated metabolites respectively which were identified by matching standards or the MS/MS spectra of the compounds. At this stage, there was no coincidence among platforms regarding the suggested metabolites. When screened on the raw data, 6 and 5 of these compounds were detected on the other platform with a similar trend. Some of the detected metabolites showed complimentary information when integrated on biological pathways. Through the use of some examples at the annotated metabolite level, possible explanations of this initial divergence on the results are presented. This work shows the complications that may arise on the comparison of non-targeted metabolomics platforms even when metabolite focused approaches are used in the identification.
Keywords: LC-HRMS; Metabolite identification; Metabolomics; Wine.
Copyright © 2016 Elsevier B.V. All rights reserved.