Abnormal savda is a special symptom in Uigur medicine. The understanding of its metabolic origins is of great importance for the subsequent treatment. Here, a metabonomic study of this symptom was carried out using LC-MS based human serum metabolic profiling. Orthogonal signal correction partial least-squares discriminant analysis (OSC-PLS-DA) was used for the classification and prediction of abnormal savda. Potential biomarkers from metabonomics were also identified for a metabolic understanding of abnormal savda. As a result, our OSC-PLS-DA model had a satisfactory ability for separation and prediction of abnormal savda. The potential biomarkers including bilirubin, bile acids, tryptophan, phenylalanine and lyso-phosphatidylcholines indicated that abnormal savda could be related to some abnormal metabolisms within the body, including energy metabolism, absorption of nutrition, metabolism of lecithin on cell membrane, etc. To the best of our knowledge, this is the first study of abnormal savda based on serum metabolic profiling. The LC/MS-based metabonomic platform could be a powerful tool for the classification of symptoms and for the development of this traditional medicine into an evidence-based one.