Phenotyping human blood plasma by 1H-NMR: a robust protocol based on metabolite spiking and its evaluation in breast cancer

This study reports an accurate assignment of the resonance signals present in H-1-NMR spectra of human blood plasma. Hereto, blood plasma was spiked with 32 different metabolites in relevant concentrations since reported chemical shift values show quite some variability depending on the biofluid under study and the applied experimental measuring conditions. The resulting information was used to rationally divide the H-1-NMR spectrum in 110 well-defined integration regions for application in metabolomics. A case-control dataset of 53 breast cancer patients and 52 controls was investigated in order to demonstrate the proof of principle. After removal of noisy variables, i.e. variables exceeding a premised threshold for the coefficient of variation, the groups could be discriminated by OPLS-DA multivariate statistics with a sensitivity and specificity of 83 and 94 %, respectively. In addition, the classification was validated in a small but independent cohort. The proposed methodology might pave the way towards a better understanding of disturbances in disease-related biochemical pathways and so, to the clinical relevance of study findings.