0 avis
Contribution of an integrative multi-omic approach in the metabolic syndrome prediction: a nested case-control study. Systems Medicine, personalised health and therapy
Archive ouverte
Edité par CCSD -
Acknowledgments: Project supported by the Fondation Francophone de Recherche sur le Diabète. International audience. Background: The rising worldwide prevalence of metabolic syndrome (MetS), a cluster of cardiometabolic risk factors of predictive of type 2 diabetes, relates largely to increasing obesity and sedentary but also to early metabolic life events. Objective: The objective of the study was to identify predictive biomarkers of evolution toward MetS 8 years later, and to bring new knowledge about this pathological state using a multidisciplinary approach in an at-risk population (subjects with small birth weight). Design: This case-control study (subjects free of MetS at baseline (n=92 born small vs n=76 born adequate for gestational age (SGA vs SGA)) was nested in the French community-based Haguenau cohort. The control group was randomly matched for age and sex. Serum signatures were determined and compared at baseline (20 years old) to determine predictive biomarkers using both untargeted mass spectrometry metabolomics and targeted proteomics using microarrays. Individual predictive models were first built using linear logistic regressions from the omics datasets. Metabolomic and proteomic data were finally integrated using random forest to determine whether multidimensional models improve prediction. Results: Univariate statistical analyses allowed identifying 93 discriminant metabolites and 47 proteins between cases and controls at baseline, with in both cases, specific gender differences. The resulting models based on either 4 metabolites or 4 proteins showed good performances: 22% misclassification on training set, 25% on validation set vs 11% misclassification on training set, 33% on validation set, respectively. Multi-omic data integration improved performance and robustness of the prediction (11% misclassification on training set, 8% on validation set). Correlation analyses with other data (anthropometric, biochemical) contributed to better understand the role of these biomarkers in the pathological processes, and therefore to evaluate their potential clinical value. Conclusion: These results should provide new tools to better stratify at-risk populations, and additional knowledge on MetS development
Consulter en ligne
Suggestions
Du même auteur
