Promotor: T. Hankemeier, Co-promotor: T.H. Reijmers
|Auteur||F.M. van der Kloet|
|Links||Thesis in Leiden Repository|
The aim of this thesis was to develop concepts and methods to extract qualitative and quantitative information about metabolites from untargeted mass spectrometric data of biological samples. Several typical challenges in data handling were addressed that prevent a straightforward interpretation (data analysis) of the data acquired with different types of mass spectrometric-based metabolomics methods (GC-MS, LC-MS, CE-MS or DI-MS) methods. The critical parameters causing variation in quantitative results were identified and studied at different stages in the metabolomics workflow such as data acquisition, data pre-processing and data analysis. Different methods and concepts were developed to address these and to improve the quantitation of metabolites and the comparison between metabolite data in different samples of the same study measured at different moments or between studies. The methods developed focused on improved normalization, data pre-processing of untargeted analysis and data pre-processing of high resolution direct infusion mass spectrometry data. Furthermore it was demonstrated that even for metabolomic studies with few samples cross-validation of multivariate models can be very time consuming and parallel implementation on a (large) cluster of computers is the way to make such computations feasible.