Résumé
A national project was conducted within the MetaboHub infrastructure to investigate thecomplementarity and redundancy of metabolomics and lipidomics analytical methods, applied on thep53 tumor suppressor and several of its key regulators. The project involved untargeted NMRapproaches, especially fast 2D NMR spectra for enhanced signal separation and integration comparedto 1H NMR. The assessment of their statistical performances and their advantages in terms ofannotation is explored.Lyophilized livers of 112 mice (fed or fasted) harvested from liver-specific conditional knock-out micefor p53, Mdm2 and E4f1 (eight genotypes) were extracted using the Bligh and Dyer protocol. 1H andNUS-TOCSY experiments were conducted on the metabolomic extracts while 1H and fast 2D (UF-COSY,NUS-TOCSY, NUS-HSQC) were performed on the lipid extracts. 1D spectra were processed usingTopSpin and NMRProcFlow, while various software were tested for 2D datasets, including DeepPicker,PeakViewer and APPIN. Exhaustive annotation was performed by leveraging the complementaritybetween the 1D and 2D datasets. Then, statistical analyses were conducted using non-supervised(ComDim) and supervised (ConsensusOPLS) multiblock approaches.Multiblock analyses revealed better statistical performances by integrating the six datasets together.These analyses distinguished fed from fasted groups, but not genotypes. Lipidomic datasetscontributed more than metabolomics ones to this separation, showing that lipids are more impactedby the mice’s diet than polar metabolites. For lipidomics, NUS-TOCSY and UF COSY demonstratedsimilar contributions to the model as the 1D dataset, while the use of all 2D datasets allowed theannotation of 53 lipid signals with increased confidence.