The challenge to produce robust data in UHPLC-HRMS/MS based metabolomics for molecular epidemiology studies

International audience ; Metabolomics studies can be a powerful tool to identify biomarkers of exposures and/or effects. The aim of this project is to develop a general method for the evaluation of agricultural workers exposome through UHPLC-HRMS/MS metabolomic to obtain robust and meaningful results.Each step of the workflow must be optimized. The optimizable steps are sample preparation, chromatographic separation, high-resolution mass spectrometry analysis, data pre-treatment, statistical analysis and biological interpretation of the identified biomarkers. Among these steps analytical measurement is critical to collect as much reliable data as possible while being relevant.This step was optimised on urine collected from a group of healthy volunteers composed of 11 men and 10 women. A set of 5 mobile phases usually used for UHPLC in the literature were tested with a QTOF detector. Samples were injected one or several times to estimate the importance of injection replication. Three normalisation methods were tested, based on creatinine, urine osmolarity and total MS signal. Two ionisation modes were used, combined with reverse phase and HILIC separations. The association of these conditions were evaluated using Galaxy Workflow4Metabolomics. The results were evaluated according to the total numbers of features, and results repeatability. Several data treatment methods were then tested with specifics parameters. At the end, a separation of men and women in 2 groups have been obtained using univariate and multivariate statistical analysis.This method will be applied to the analysis of biological samples obtained from large agricultural occupational cohorts. Other steps still need to be optimized before the global can be applied to cohort analysis. The data treatment step can be improved by the utilisation of external compounds or endogens molecules as standards for the retention time correction of the detected peaks. Statistical treatment step may be more reliable by using new statistical. Biomarkers ...