A Study of spectral integration and normalization in NMR-based metabonomic analyses
- 作者：Bobbie-Jo M. Webb-Robertson ; David F. Lowry ; Kristin H. Jarman ; Sam J. Harbo ; Q. Ryan Meng ; Alfred F. Fuciarelli ; Joel G. Pounds and K. Monica Lee
- 关键词：Metabonomics ; Nuclear magnetic resonance (NMR) ; Principal component analysis (PCA) ; Normalization ; Integration ; Cross-validation
- 刊名：Journal of Pharmaceutical and Biomedical Analysis
- 出版年：2005
- 期刊代码：65_07317085
- 类别：ch
- 出版时间：2005
- 卷：39
- 期：3-4
- 页码：830-836
- 文件大小：268 K
摘要
Metabonomics involves the quantitation of the dynamic multivariate metabolic response of an organism to a pathological event or genetic modification [J.K. Nicholson, J.C. Lindon, E. Holmes, Xenobiotica 29 (1999) 1181–1189]. The analysis of these data involves the use of appropriate multivariate statistical methods; Principal Component Analysis (PCA) has been documented as a valuable pattern recognition technique for 1H NMR spectral data [J.T. Brindle, H. Antti, E. Holmes, G. Tranter, J.K. Nicholson, H.W. Bethell, S. Clarke, P.M. Schofield, E. McKilligin, D.E. Mosedale, D.J. Grainger, Nat. Med. 8 (2002) 1439–1444; B.C. Potts, A.J. Deese, G.J. Stevens, M.D. Reily, D.G. Robertson, J. Theiss, J. Pharm. Biomed. Anal. 26 (2001) 463–476; D.G. Robertson, M.D. Reily, R.E. Sigler, D.F. Wells, D.A. Paterson, T.K. Braden, Toxicol. Sci. 57 (2000) 326–337; L.C. Robosky, D.G. Robertson, J.D. Baker, S. Rane, M.D. Reily, Comb. Chem. High Throughput Screen. 5 (2002) 651–662]. Prior to PCA the raw data is typically processed through four steps; (1) baseline correction, (2) endogenous peak removal, (3) integration over spectral regions to reduce the number of variables, and (4) normalization. The effect of the size of spectral integration regions and normalization has not been well studied. The variability structure and classification accuracy on two distinctly different datasets are assessed via PCA and a leave-one-out cross-validation approach under two normalization approaches and an array of spectral integration regions. The first dataset consists of urine from 15 male Wistar–Hannover rats dosed with ANIT measured at five time points, mimicking drug-induced cholangiolitic hepatitis [D.G. Robertson, M.D. Reily, R.E. Sigler, D.F. Wells, D.A. Paterson, T.K. Braden, Toxicol. Sci. 57 (2000) 326–337; J.P. Shockcor, E. Holmes, Curr. Top. Med. Chem. 2 (2002) 35–51; N.J. Waters, E. Holmes, A. Williams, C.J. Waterfield, R.D. Farrant, J.K. Nicholson, Chem. Res. Toxicol. 14 (2001) 1401–1412]. The second data is serum samples from young male C57BL/6 mice subjected to instillation of pancreatic elastase producing emphysema type symptoms [C. Kuhn, S.Y. Yu, M. Chraplyvy, H.E. Linder, R.M. Senior, Lab. Invest. 34 (1976) 372–380; C. Kuhn, R.M. Senior, Lung 155 (1978) 185–197]. This study indicates that independent of the normalization method the classification accuracy achieved from metabonomic studies is not highly sensitive to the size of the spectral integration region. Additionally, both datasets scaled to mean zero and unity variance (auto-scaled) have higher variability within classification accuracy over spectral integration window widths than data scaled to the total intensity of the spectrum. Of the top 10 latent variables for the ANIT dataset the auto-scale normalization has standard deviations larger than the total-scale in seven cases. In the case of the elastase all standard deviations are larger for the auto-scaling.