Profinder software b 08

All data were controlled and acquired using Mass Hunter Qualitative Analysis B.07.00 (Agilent Technologies). Data obtained from LC-MS and CE-MS were cleaned of background noise, unrelated ions, peak detection, deconvolution and alignment were performed by the recursive feature extraction (RFE) using Profinder Software B.08.00 (Agilent Technologies). Data obtained from GC-MS were processed by Unknowns Analysis Software B.09.00 (Agilent Technologies) to perform peak extraction and deconvolution. Alignment and integration were performed by Mass Profiler Professional 14.9 and MS Quantitative Analysis (Quant-My-Way) B.09.00. Blank subtraction and filtering by frequency of at least 50% of the QC and 70% of each group and relative standard deviation (RSD) less than 20% (LC-MS) and 30% (CE-MS and GC-MS) in QC were performed, to keep only the relevant features. Feature with higher RSD was removed from the data set. Missing values were substituted by KNN algorithm using MetaboAnalyst 4.0 Software [23 (link)]. Normalization by QC was applied to all data matrix and I.S applied only to GC-MS data. Beyond this, all data were log-transformed to get a normal distribution and used pareto scaling.