“…The following settings were employed to preprocess the data using Mzmine TM (BMC Bioinformatics, United Kingdom): mass detection using exact mass algorithm (noise level, 1.0E6); chromatogram builder (minimum time span in min, 0.2; minimum height, 3.0E6; m/z tolerance, 0.001 m/z or 5.0 ppm); chromatogram deconvolution using baseline cut-off (minimum peak height, 3.0E6; peak duration, 0.2 to 5.0; baseline level, 1.0E6); isotopic peaks grouper (m/z tolerance, 0.002 m/z or 5 ppm; retention time tolerance in min, 0.2; maximum charge, 2; representative isotope, most intense); gap filling (intensity tolerance, 30%; m/z tolerance, 0.001 m/z or 5.0 ppm; retention time tolerance in min, 0.5) and alignment using join aligner (m/z tolerance, 0.001 m/z or 5 ppm; weight for m/z, 20; retention time tolerance, 5%; weight for retention time, 20). The data matrices of both positive and negative ionization modes were subsequently combined on one spreadsheet, after preprocessing individually, and prior to multivariate statistical analysis, the generated data matrix was subjected to log10transformation ( VAN DEN BERG et al, 2006;BALLABIO;CONSONNI, 2013;OVENDEN et al, 2014). Unsupervised (HCA and PCA) and supervised (PLS-DA) statistical analysis, using the software SIMCA 13.0.3.0© (Umetrics, Sweden), were used to perform data mining.…”