We aligned and normalized the raw data, then excluded the
values of RSD 4 20% within one group. Finally, a total of 896 peaks in ESI þ and 730 peaks in ESI were merged and imported into the SIMCA-P software for multivariate statistical analysis (Smilde et al., 2005). To investigate the global metabolism variations, we first used PCA to analyze all observations acquired in both ion modes. PCA, as an unsupervised pattern recognition method for handling metabolomics data, can classify the metabolic pheno- types based on all imported samples. As shown in the PCA score
3D plot (Fig. 3), an overview of all samples in the data can be observed and exhibit a clear grouping trend (R2X, 0.503; Q2, 0.298) between sham, model and BNC-treated groups. This observation indicates that MCAO disturbs metabolism of endogenous sub- stances compared to the normal state. Furthermore, low-level doses of BNC had an effect on MCAO-induced ischemia, although the trajectory of the BNC treated groups did not return to baseline values.