Comprehensive two-dimensional gas chromatography/time-of-flight mass spectrometry (GCxGC-TOFMS) coupled with pattern recognition methods was applied to analyze plasma from diabetic patients and healthy controls. After sample preparation and GCxGC-TOFMS analysis, collected data were transformed, the peak alignment between different chromatograms was performed to generate the metabolites' peak table, then orthogonal signal correction filtered partial least-squares discriminant analysis (OSC-PLSDA) was carried out to model the data and discover metabolites with a significant concentration change in diabetic patients. With the method above, diabetic patients and healthy controls could be correctly distinguished based on the metabolic abnormity in plasma. Five potential biomarkers including glucose, 2-hydroxyisobutyric acid, linoleic acid, palmitic acid and phosphate were identified. It was found that elevated free fatty acids were essential pathophysiological factors in diabetes mellitus which reflected either the hyperglycemia or the deregulation of fatty acids metabolism. These potential biomarkers in plasma, e.g. palmitic acid, linoleic acid and 2-hydroxybutyric acid might be helpful in the diagnosis or further study of diabetes mellitus. This study shows the practicability and advantage of GCxGC-TOFMS coupled with data analysis and mining for metabonomics in biomarker discovery.