Abstract: Diabetes is a common metabolic syndrome which presents a serious threat to human health. Traditional Chinese medicine (TCM) has been widely paid attention to its advantages and characteristics in the diagnosis and the treatment of diabetes. A strategy of classifying five TCM syndromes in diabetes (Qi-deficiency, Yin-deficiency, Qi- and Yin-deficiency, Damp heat and Blood stasis) was employed based on plasma fatty acid metabolic profiles, lipid metabolism indicators and chemometrics methods. Using orthogonal signal correction-partial least squares (OSC-PLS) method, the five syndromes were obviously distinguished from those of the health control, which confirmed there existed metabolite differences in different traditional Chinese medicine syndromes. Furthermore, a new method, uncorrelated linear discriminant analysis (ULDA), was applied in the discrimination of health control, TCM deficiency syndromes (Qi-deficiency, Yin-deficiency, Qi- and Yin-deficiency) and TCM empirical syndromes (Damp heat, Blood stasis), which demonstrated better clustering results, the correct rate reached 95.7%. The four potential biomarkers, C20:2, C20:5, triglycerides (TG) and high density lipoprotein (HDL), performed large contributions to the classification which can provide important information assisting TCM clinical diagnosis.

Key words: diabetes, syndrome classification, uncorrelated linear discriminant analysis (ULDA), fatty acid metabolic profiles