Metabolomic Investigation Of Three Complex Biosvstems By Using NMR Approach

Metabolomics is a developing discipline targeted to explore the essence of life process through investigation of overall changes of small molecular compounds in complex biosystems including biofluids, tissues and organs, which contains great varieties of proton-rich metabolites. NMR is one of the most commonly used detection technology in metabolomics due to its many advantages such as simple sample handling, abundant spectra editing methods, unbiased detection, and excellent reproducibility. Thus,1H NMR-based metabolomics is a powerful tool to investigate complex biosystems with the obtained raw data being a set of NMR spectra. Before depth data mining of the NMR spectra using chemometric models, it is necessary to firstly reduce the data set to a manageable proportion to facilitate subsequent statistical analysis. Currently, spectral binning is the most widely used NMR data reduction method which first divide the original spectra into a number of buckets with a width of about 2 Hz and then get the integration of peak intensities within each bucket to obtain the integrated spectra. Spectral binning is an efficient and easily automated data reduction method and is particularly suitable for large-scale sample studies. The major drawbacks of spectral binning are that there are too many variables in the binned data set resulting in dramatically declined performance of chemometric models and the reduction of spectral resolution will sometimes hamper the assignment of characteristic variables. When the sample size is small, we can also employ the spectral deconvolution method to reduce the NMR spectra data set. During spectral deconvolution the experimental spectrum is fitted by standard profiles of various metabolites, resulting in a very few variables. The disadvantages of this method are that it is highly dependent on metabolite profiles database and difficult to be automated. Therefore, spectral deconvolution is only suitable for early exploratory investigations which usually involve just a small of number of samples.In chapter 2, we assessed the impact of the two NMR data reduction methods mentioned above on the performance of principal component analysis (PCA) and orthogonal projection to latent structures (OPLS) models, and then evaluated the metabolic changes in the spleen and stomach of C57BL/6J male mice after metastasis of primary melanoma cells implanted at the flank location on the legs of a B16-F10 melanoma mouse model. Chemomentric data analysis revealed that the melanoma group was well separated from the control group and metastatic melanoma may have seriously disturbed multiple biological pathways leading to significant changes of metabolic phenotypes in the spleen and stomach tissue. In chapter 3, we employed the same research strategy as used in chapter 2 to investigate the metabolite signatures in the lungs of both regular and obese C57BL/6J mice with and without being exposed to a controlled amount of cigarette smoke. OPLS modeling demonstrated that the concentrations of adenosine derivatives, i.e. ATP, ADP, and AMP, were significantly decreased, while the levels of inosine and uridine were apparently elevated in the lungs of mice exposed to cigarette smoke when compared with controls regardless the mice were obese or of regular weight. We also found that the ratio of glycerophosphocholine (GPC) to phosphocholine (PC) was significantly increased in the lungs of obese mice compared with those of the regular weight mice, while cigarette smoke exposure alone cannot make any obvious alteration to the GPC/PC ratio. However, the GPC/PC ratio was further elevated in the lungs of obese mice exposed to cigarette smoke.We focused on investigating variances between different groups of samples in chaper 2 and chapter 3, while ignoring individual differences within each group. However, individual difference is also very important since it can reflect the varied responses of different individuals to the same external stimulus. Thus, in chapter 4, we studied the individual diversities of human plasma-ibuprofen (IBP) interaction. This chapter is a complement and extension of a previous work published by our group. In this chapter, the diversity index (Idiv) and the interaction index (Idist) were redefined to improve the analysis and to facilitate the interpretation of the results. Multivariate data analysis showed that the redefined Idiv and Idist were superior to the previously defined ones with regard to characterizing the sample diversity and individual differences of IBP-induced metabolic profile changes, respectively. Also, the redefined Idist can extract more information from the NMR spectra and the redefined Idiv was uniquely determined compared with the previously defined ones, respectively. More importantly, the redefined Idist and Idiv showed extremely significant linear correlationship (p< 2.2e-16). As a complementary work of our previous publication, we analyzed the correlationship between Idist and NMR spectra, Idist and clinical data, clinical data and NMR spectra, respectively, and found that the diversity of IBP-plasma interaction was associated with individual differences of binding strength between IBP as well as small molecular weight metabolites and lipoproteins, respectively. We also found that the higher levels of free lactate or the lower concentrations of total triglycerides, albumin or choline in human plasma, or the lighter weight, the smaller body mass index (BMI), the more sensitive to IBP. The older volunteers recruited in the present study, the higher levels of glucose and glycoproteins, as well as the lower levels of tyrosine, valine, creatinine and choline-containing phospholipids in lipoprotein particles were observed in plasma.In the meantime, we can extract metabolites showing signals in the NMR spectra that were in the same or nearby pathways as that of the targeted metabolites from the correlation coefficients coded OPLS loadings plots.