Analyzing Metabolism with NMR Spectroscopy: A Comparison of Multivariate Statistical Tools in MetaboAnalyst, MATLAB, and R

Abstract

This thesis compares the multivariate statistical analysis capabilities of MetaboAnalyst51, MATLAB52, and R53 for NMR metabolomics data. A comprehensive assessment was conducted using a dataset of high-resolution 1H NMR data from 40 horse serum samples. The aim was to evaluate the analytical accuracy, computational efficiency, and STOCSY capability of each software platform. The results showed that MetaboAnalyst51 is the most user-friendly and computationally efficient platform, while R53 provides the best documentation and example outputs for STOCSY analysis. MATLAB52 excels in handling resource-intensive tasks and large datasets. Recommendations were provided for researchers with different programming skills and familiarity with the platforms. The study highlights the need for standardized approaches and further optimization in metabolomics research. Future directions include incorporating diffusion filtered data, investigating persistent outliers, and exploring other software platforms. Overall, metabolic phenotyping holds promise for transforming the field of medical diagnosis.