Application of several pattern recognition methods in metabonomic research of rat hyperlipidemia

Aim: To investigate the application of several pattern recognition methods in metabonomic research of rat hyperlipide. Methods: The hyperlipidemia rat model was developed successfully by feeding with high fat diet for 4 weeks. During this period, the course of disease was studied by GC-MS-based metabonomic approach, and there are some endogenous metabolites identified in the plasma sample by chromatogram. Then, these endogenous metabolites were processed by three different pattern recognitions respectively, including principle components analysis (PCA), soft independent modeling of class analogy (SMICA), and non-linear mapping (NLM). Results: The results showed that the clustering power of SIMCA and NLM is superior to that of PCA, and the data processed by NLM can reflect the track of diverse stages of hyperlipidemia development in rats, while PCA and SMICA could elucidate the endogenous metabolites related to hyperlipidemia and help to find potential biomarkers. Conclusion: The outcomes from three methods were complementary and demonstrated that various pattern recognitions could be used in metabonomic study to get more reliable and comprehensive information.