ABSTRACT
This is a report on how 1H NMR-based metabonomics was employed to discriminate osteopenia from osteoporosis in postmenopausal women, identifying the main metabolites associated to the separation between the groups. The Assays were performed using seventy-eight samples, being twenty-eight healthy volunteers, twenty-six osteopenia patients and twenty-four osteoporosis patients. PCA, LDA, PLS-DA and OPLS-DA formalisms were used. PCA discriminated the samples from healthy volunteers from diseased patient samples. Osteopenia-osteoporosis discrimination was only obtained using Analysis Discriminants formalisms, as LDA, PLS-DA and OPLS-DA. The metabonomics model using LDA formalism presented 88.0% accuracy, 88.5% specificity and 88.0% sensitivity. Cross-Validation, however, presented some problems as the accuracy of modeling decreased. LOOCV resulted in 78.0% accuracy. The OPLS-DA based model was better: R2Y and Q2 values equal to 0.871 (p<0.001) and 0.415 (p<0.001). LDA and OPLS-DA indicated the important spectral regions for discrimination, making possible to assign the metabolites involved in the skeletal system homeostasis, as follows: VLDL, LDL, leucine, isoleucine, allantoin, taurine and unsaturated lipids. These results indicate that 1H NMR-based metabonomics can be used as a diagnosis tool to discriminate osteoporosis from osteopenia using a single serum sample.