Profiler version 8

Variables in the OPLS-DA model with both values of variable importance in projection of greater than one (VIP > 1) and error bar not exceeding zero were considered as discriminant metabolites. Characterization of discriminant metabolites was carried out based on the match between experimental ¹H-NMR signals and reference compound of in-house library of Chenomx Profiler version 8.2 (Chenomx Inc., Edmonton, AB, Canada).
The identity of characterized metabolites was further supported by comparison of two-dimensional (2D) ¹H-¹H J-resolved and heteronuclear single quantum correlation (HSQC) experiments with online databases such as the Human Metabolome Database (HMDB), Kyoto Encyclopedia of Genes and Genomes (KEGG), PubChem, and ChemSpider.

NMR spectra of all samples were superimposed to identify meaningful regions (δ 0.50–9.70) and residual solvent signals (methanol: δ 3.27–3.33 and water: δ 4.70–5.07) using MestReNova version 6.0.2-5475 (Mestrelab Research S.L., Santiago de Compostela, Spain). The NMR spectra were further processed using Chenomx Processor version 8.2 (Chenomx, Edmonton, AB, Canada) with the following parameters: chemical shape indicator (CSI) = TSP, TSP concentration = 0.2902 mM, sample pH = 6.0 ± 0.5, and automatic phase, baseline (cubic spline), and shim correction. Identification and quantification of metabolites were carried out by fitting the NMR spectra to reference spectra in Chenomx Spectral Reference Library in Chenomx Profiler version 8.2. Identity and estimated concentration of a metabolite was considered valid only in case of all NMR signals of the metabolite fitting perfectly to all signals of the metabolite in the reference library and no overlapping with signals of other metabolites. The NMR spectra were normalized to total area and the residual solvent signals were removed. Next, the meaningful regions were discretized to buckets of δ 0.04 width, producing a data matrix dimension of 18 × 221 in tsv file format.