This paper presents an liquid chromatography (LC)/mass spectrometry (MS)-based metabonomic platform that combined the discovery of differential metabolites through principal component analysis (PCA) with the verification by selective multiple reaction monitoring (MRM). These methods were applied to analyze plasma samples from liver disease patients and healthy donors. LC-MS raw data (about 1000 compounds), from the plasma of liver failure patients (n = 26) and healthy controls (n = 16), were analyzed through the PCA method and a pattern recognition profile that had significant difference between liver failure patients and healthy controls (P < 0.05) was established. The profile was verified in 165 clinical subjects. The specificity and sensitivity of this model in predicting liver failure were 94.3 and 100.0%, respectively. The differential ions with m/z of 414.5, 432.0, 520.5, and 775.0 were verified to be consistent with the results from PCA by MRM mode in 40 clinical samples, and were proved not to be caused by the medicines taken by patients through rat model experiments. The compound with m/z of 520.5 was identified to be 1-Linoleoylglycerophosphocholine or 1-Linoleoylphosphatidylcholine through exact mass measurements performed using Ion Trap-Time-of-Flight MS and METLIN Metabolite Database search. In all, it was the first time to integrate metabonomic study and MRM relative quantification of differential peaks in a large number of clinical samples. Thereafter, a rat model was used to exclude drug effects on the abundance of differential ion peaks. 1-Linoleoylglycerophosphocholine or 1-Linoleoylphosphatidylcholine, a potential biomarker, was identified. The LC/MS-based metabonomic platform could be a powerful tool for the metabonomic screening of plasma biomarkers.