Fig. 2

RPA administration induced changes in the serum metabolism profile in mice. a Feature extraction analyses based on six data dimension-reduction algorithms (n = 7 CON group and 6 RPA group samples). b ROC curves based on the CV performance by the SVM classifier, the default is the ROC curves from all models averaged from all CV runs, and the 95% confidence interval can be computed in this case. c Model of 20 features that were selected as screening criteria, AUC = 0.94. d Correlation analysis network among all identified serum metabolites. e Differential metabolite analysis model between groups based on OPLS-DA. f Pathways enriched by differential metabolites-based KEGG; the redder the color is, the more significant the result. p < 0.05. PCA principal component analysis, t-SNE T-distributed stochastic neighbor embedding, ISOMAP isometric mapping, LLE locally linear embedding, WT wavelet transform, ROC receiver operating characteristic, CV cross-validation, AUC area under the curve