annual plant reviews volume 43 (biology of plant metabolomics) progress in chemometrics and biostatistics for plant applications, or a good red wine is a bad white wine推荐.pdfVIP

Annual Plant Reviews (2011) 43, 317–342 doi: 10.1002/9781444339956.ch10 Chapter 10 PROGRESS IN CHEMOMETRICS AND BIOSTATISTICS FOR PLANT APPLICATIONS, OR: A GOOD RED WINE IS A BAD WHITE WINE 1 1 2 Joachim Kopka , Dirk Walther , J. William Allwood and Royston Goodacre2 1 ¨ Max Planck Institute for Molecular Plant Physiology, Am Muhlenberg 1, 14476 Golm, Germany 2 School of Chemistry, Manchester Interdisciplinary Biocentre, University of Manchester, 131 Princess Street, Manchester, M1 7DN, UK Abstract: In the hunt for novel chemometric and biostatistical tools suitable for the automated mining of large metabolite profiling and fingerprinting data sets, the application of well-established statistical tools appears to have been neglected. Using a data set that aims to understand the contributions that the metabolome can make to the prediction of the enological quality of white wines from the Palati- nate region of Germany, we exemplify the power of ‘simple’ statistical analyses for knowledge generation and repeatable data reduction without loss of relevant information. In view of the many and diverse experimental designs required in plant physiological studies, we emphasize that no generalized standard operation procedure for the statistical data mining of such experiments should be suggested. Instead, a well-compiled toolbox of simple and enhanced statistical algorithms should be at hand, and this chemometrics ‘zoo’ can be applied in a case-by-case decision and sequence. Care should be taken to have more confidence in statistical significance analyses that estimate error probabilities rather than to focus on just a few sample classification tools, such as the omnipresent principal component (PCA) and hierarchical cluster analyses (HCA). These t