Multivariate data analysis was implemented using the default parameters on the SIMCA 14 software (Umetrics, Umeå, Sweden). The score plots of orthogonal projections to latent structures dis- criminant analysis (OPLS-DA) model was generated on the posi- tive- and negative-mode MS data obtained from the XCMS pro- gram (Section 2.4), which had been
centered and scaled to Pareto variance.