To identify the differential metabolites of CRC, the metabolomes of tumor tissues were compared with that of matched adjacent mucosa. Supplementary Table 1 shows the demographic characteristics and clinicopathological features of 51 CRC patients. Mass spectrometry detected 4,526 and 4,765 variables in negative electrospray ionization (ESI-) and positive electrospray ionization (ESI+), respectively. Multivariate analysis was performed on the result of mass spectrometry to find metabolites that mostly discriminated the study groups. Principal component analysis (PCA) was the unsupervised analysis method, which was used for dimension reduction of data through making a linear combination of variables known as principal components. PCA analysis can reveal trends in the data and groups of observations and find outliers. Although a weak trend in clustering according to the PCA plot based on tumor tissue and adjacent mucosa was observed, the PCA analysis results showed a separation of tumor tissue and adjacent mucosa into two clusters (Figures 1A,D). To further study the differences between tumor tissue and adjacent mucosa and to find potential biomarkers, the supervised multivariate statistical method OPLSDA was subsequently used. OPLS-DA is a supervised analysis method that is employed to divide the samples into different groups, including tumor tissue and adjacent mucosa, which was performed to find metabolites that mostly discriminated the studied groups in each comparison. The classification results are shown in Figures 1B,E. To guard against model overfitting, permutation tests (100 random permutations) were performed. These permutation tests were used to contrast the goodness of fit of the original model with the goodness of fit of randomly permuted models. As shown in Figures 1C,F, the validation plots strongly indicated that the original combined models were valid. No overfitting was observed. A total of 373 metabolites (296 higher and 77 lower) were identified with the criteria of Variable important for the projection (VIP) score >1.5 and P-values of