Objective To investigate and characterize the serum metabolite profile of patients with rheumatoid arthritis (RA),a common autoimmune disease,which will be of help for early diagnosis in clinic.Methods Serum specimens from 26 patients with RA and 19 age-matched healthy volunteers were collected from the First Affiliated Hospital of Hunan University of Chinese Medicine from January through April 2015.Samples were detected on the gas chromatography-mass spectrometry (GC-MS),and serum metabolites were identified by the chemometric methods.Discriminative model of RA and healthy volunteers was established using partial least squares-linear discriminant analysis (PLS-LDA),and furthermore the established model was evaluated with double cross validation (DCV) for predicting ability.Finally,differential metabolites with clinical diagnostic potential were screened out by using subwindow permutation analysis (SPA).Results The established PLS-LDA discriminative model identified 48 metabolites.The total accuracy of the model approached 97.73％,in which the accuracy of the model for predicting RA and healthy volunteers was 96.15％ and 100％,respectively.Our studies screened out 3-hydroxy butyric acid,phosphoric acid,isoleucine,mannose and hexadecanoic acid with clinical diagnostic potential based on the SPA.Conclusions Metabolomics with the application of GC-MS combined with the PLS-LDA method can distinguish patients with RA and healthy volunteers,and can aid in potential early diagnosis of RA.