conventional biochemical markers have limited usefulness in the prediction of early liver dysfunction. We, therefore, tried to find more useful liver failure biomarkers after liver resection that are highly sensitive to internal and external challenges in the biological system with a focus on liver metabolites. Twenty pigs were divided into the following 3 groups: sham operation group (n = 6), 70% hepatectomy group (n = 7) as a safety margin of resection model, and 90% hepatectomy group (n = 7) as a liver failure model. Blood sampling was performed preoperatively and at 1, 6, 14, 30, 38, and 48 hours after surgery, and 129 primary metabolites were profiled. Orthogonal projection to latent structures-discriminant analysis revealed that, unlike in the 70% hepatectomy and sham operation groups, central carbon metabolism was the most significant factor in the 90% hepatectomy group. Binary logistic regression analysis was used to develop a predictive model for mortality risk following hepatectomy. the recommended variables were malic acid, methionine, tryptophan, glucose, and γ-aminobutyric acid. Area under the curve of the linear combination of five metabolites was 0.993 (95% confidence interval: 0.927-1.000, sensitivity: 100.0, specificity: 94.87). We proposed robust biomarker panels that can accurately predict mortality risk associated with hepatectomy. Liver resection is inevitable in the treatment of specific liver diseases. However, reduced liver volume can have many clinical manifestations such as small-for-size syndrome and post-hepatectomy liver failure, especially in cases of large volume resections of the liver. One of the main reasons for deteriorating liver function is high portal pressure, which results in high shear stress and sinusoidal injury 1. Conventional biochemical parameters for the evaluation of liver function in clinical settings, such as total bilirubin and prothrombin time (PT), are used as predictive markers. However, these biomarkers have limited usefulness in the prediction of early dysfunction. Total bilirubin and PT only have a 59% sensitivity in the prediction of postoperative mortality 2. Therefore, there is a need for more specific and accurate predictive markers. Metabolites are the final readouts of a range of endogenous biochemical activities. Changes in the metabolic profile are highly sensitive to internal and external challenges in biological systems. Thus, comprehensive metabolome information has great value in the diagnosis and prognosis of various diseases. In this study, we used liver failure animal models to determine changes in the levels of serum metabolites that can be identified before conventional biochemical markers. Two pig models were used to investigate the regenerating or failing liver after liver resection. Resection of 70% of the liver provided a model of liver injuries that did not lead to liver failure and served as a successful regenerative model, and resection of 90% of the liver provided a model of liver failure that eventually led to death. We identified ...