Objective To identify potential biomarkers and establish a prediction model for chemotherapy-related hepatotoxicity susceptibility based on plasma endogenous metabolites of colorectal cancer (CRC) patients before chemotherapy. Methods The plasma samples of 50 CRC patients before capecitabine chemotherapy and the records of their chemotherapy-related hepatotoxicity during the follow-up were collected. An ultra-high-performance liquid chromatography coupled to quadrupole time-of-flight mass spectrometry (UHPLC-Q-TOF-MS) was used to perform untargeted metabolomic analysis. Based on bioinformatics analysis, differential analysis, correlation analysis, and random forest were used to screen for hepatotoxicity-related plasma endogenous metabolites. All samples were randomly assigned (7∶3) to training set or test set. A multivariate logistic regression model was established to predict the hepatotoxicity of capecitabine chemotherapy based on the training set data. The prediction effects of the model in the training, test and entire sets were evaluated by receiver operating characteristic (ROC) curve analysis. Results The endogenous metabolites related to hepatotoxicity in the plasma of CRC patients before chemotherapy were mainly lipid endogenous metabolites. A series of potentially important predictive biomarkers for hepatotoxicity susceptibility were identified, including sphingamine-1-phosphate, ceramide, galactose, arachidonic acid, tyrosine, biliverdin, myristic acid, phosphatidylcholine (35∶1), phosphatidylethanolamine (36∶1), and hexadecanoic acid. The area under curve values of the prediction model based on the above biomarkers in the training, test and entire sets were 0.946 (95% confidence interval ［CI］ 0.842-1.000), 0.920 (95% CI 0.720-1.000), and 0.912 (95% CI 0.810-0.982), respectively. Conclusion The endogenous metabolites in the plasma of CRC patients before chemotherapy can effectively predict the hepatotoxicity of capecitabine chemotherapy. These hepatotoxicity biomarkers indicate that susceptible patients have characteristics related to lipid metabolism disorders