文摘
Metabonomic characterization of long-lasting althoughweak physiological events such as anabolic disruptionsremains poorly investigated. We have validated 1H-13CHMBC-NMR as a suitable generator of instrumentalvariables that are strongly linked to the concentration ofendogenous metabolites in biological fluids. This methodis interfaced to multivariate pattern recognition procedures. Fingerprints established from urine sample collected on cattle treated with anabolic steroids were usedto validate this method. Four main results arise from thisstudy. (i) 2D NMR is as informative as 1D NMR. (ii) 2DNMR variable clustering highlights successfully a contingent redundancy of variables, although a relevant hierarchical model of statistical correlations covering fromstructural relationships to physiologic links can also beevidenced. (iii) To enhance pattern recognition performances, we have validated a variable selection algorithmfor accurate prediction of unknown individuals belongingto predetermined groups achieved by linear discriminantanalysis (LDA). This algorithm synthesizes the wholeinformation contained in the data set by selecting preferentially nonredundant variables. Parameters generatingvariable subsets are validated by predicted varianceefficiency obtained when minimizing error rates calculatedby cross-validation methods. (iv) Provided variables arecorrectly filtered, LDA fairly competes with partial least-squares methods for both classification of individuals andstatistical interpretation of metabolic responses obtainedin such a physiological disruption context.