NMR-based Metabonomics' Data Preprocess Methods And Its Application On Diabetes Mellitus Study

The NMR-based metabonomics approach evolved from the pioneering work of Nicholson and co-workers in 1999 has become a novel analytical technique. NMR-based metabonomics is a systems approach for studying in vivo metabolic profiles, which can provide information on disease processes at several stages in the discovery-and-development process by detecting change of the metabolites. While, data preprocess, as normalization and scaling, is one of the key step of metabonomics analysis, there are lots of preprocess methods, however only the appropriate one can get the accurate result.Using NMR-based metabonomics as a useful analytical technique, diabetes mellitus was studied in our work. The applicability of prevalent metabonomics data normalization and scaling methods was analyzed. At the same time, variable selection methods were introduced to the data pretreatments to optimize the metabonomics data. The main results are summarized as follows:First, we studied diabetes mellitus using 1H-NMR-based metabonomics, established metabolic profiling of type 1 diabetic SD rats and identified the characteristic metabolites, also established metabolic profiling of type 2 diabetic patients and identified the characteristic metabolites.Second, we discussed the properties of different NMR-based metabonomics data normalization and scaling methods, analyzing their merits, drawbacks and applicability, analyzing their effects to the outcome of data multi-analysis. Third, we designed a fitness function based on genetic algorithm (GA), combined with parameter R which defined to evaluate the quality of PC scoring plot. GA was used to improve the data clustering quality as a variable selection method.