Cluster Analysis Statistical Spectroscopy Using Nuclear Magnetic Resonance Generated Metabolic Data Sets from Perturbed Biological Systems

详细信息    查看全文

• 作者：Steven L. Robinette ; Kirill A. Veselkov ; Eszter Bohus ; Muireann Coen ; Hector C. Keun ; Timothy M. D. Ebbels ; Olaf Beckonert ; Elaine C. Holmes ; John C. Lindon ; Jeremy K. Nicholson
• 刊名：Analytical Chemistry
• 出版年：2009
• 出版时间：August 15, 2009
• 年：2009
• 卷：81
• 期：16
• 页码：6581-6589
• 全文大小：498K
• 年卷期：v.81,no.16(August 15, 2009)
• ISSN：1520-6882

文摘

We present a new approach for analysis, information recovery, and display of biological ¹H nuclear magnetic resonance (NMR) spectral data, cluster analysis statistical spectroscopy (CLASSY), which profiles qualitative and quantitative changes in biofluid metabolic composition by utilizing a novel local−global correlation clustering scheme to identify structurally related spectral peaks and arrange metabolites by similarity of temporal dynamic variation. Underlying spectral data sets are presented in a novel graphical format to represent high-dimensionality biochemical information conveying both statistical metabolite relationships and their responses to experimental perturbation simultaneously in a high-throughput and intuitive manner. The method is exemplified using multiple 600 MHz ¹H NMR spectra of rat (n = 40) urine samples collected over 160 h following the development of experimental pancreatitis induced by l-arginine (ARG) and a wider range of model toxins including acetaminophen, galactosamine, and 2-bromoethanamine. The CLASSY approach deconvolutes complex biofluid mixture spectra into quantitative fold-change metabolic trajectories and clusters metabolites by commonalities of coexpression patterns. We demonstrate that the developing pathological processes cause coordinated changes in the levels of many compounds which share similar pathway connectivities. Variability in individual responses to toxin exposure is also readily detected and visualized allowing the assessment of interanimal variability. As an untargeted, unsupervised approach, CLASSY provides significant advantages in biological information recovery in terms of increased throughput, interpretability, and robustness and has wide potential metabonomic/metabolomic applications in clinical, toxicological, and nutritional studies of biofluids as well as in studies of cellular biochemistry, microbial fermentation monitoring, and functional genomics.