Principal component analysis (PCA) and orthogonal projections to latent structures discriminant analysis (OPLS-DA)

The progenesis QI software (Waters Corporation, Milford, USA) was employed to analyze the obtained raw date. Data processing parameters were applied as follows: precusor tolerance set to 5 ppm, fragment tolerance set to 10 ppm, and product ion threshold set to 5%. The public databases, including the Human Metabolome Database (HMDB), Lipidmaps (v2.3), and METLIN were utilized to identify metabolites. The metabolic changes across experimental groups were visualized using PCA and OPLS-DA models. By conducting a permutation test (200 times), we validated the model and minimized the risk of over-fitting. With the help of PermutMatrix73, we carried out PCA upon Pearson distances. Metabolite differences were screened based on the variable importance in projection (VIP), significance values (P values), and fold-change (FC) values. Generally, metabolites with VIP > 1 were considered to contribute significantly in discrimination interpretation, a FC threshold (FC value ≥ 1.2 or ≤ 0.83) was considered as the cutoff for up/downregulation within the concentration, and a P-value of less than 0.05 was regarded as significant.