Mélanie Pétéra

Create, run, share, publish, and reference your LC-MS, FIA-MS, GC-MS, and NMR data analysis workflows with the Workflow4Metabolomics 3.0 Galaxy online infrastructure for metabolomics

Metabolomics is a key approach in modern functional genomics and systems biology. Due to the complexity of metabolomics data, the variety of experimental designs, and the multiplicity of bioinformatics tools, providing experimenters with a simple and efficient resource to conduct comprehensive and rigorous analysis of their data is of utmost importance. In 2014, we launched the Workflow4Metabolomics (W4M; http://workflow4metabolomics.org) online infrastructure for metabolomics built on the Galaxy environment, which offers user-friendly features to build and run data analysis workflows including preprocessing, statistical analysis, and annotation steps. Here we present the new W4M 3.0 release, which contains twice as many tools as the first version, and provides two features which are, to our knowledge, unique among online resources. First, data from the four major metabolomics technologies (i.e., LC-MS, FIA-MS, GC-MS, and NMR) can be analyzed on a single platform. By using three studies in human physiology, alga evolution, and animal toxicology, we demonstrate how the 40 available tools can be easily combined to address biological issues. Second, the full analysis (including the workflow, the parameter values, the input data and output results) can be referenced with a permanent digital object identifier (DOI). Publication of data analyses is of major importance for robust and reproducible science. Furthermore, the publicly shared workflows are of high-value for e-learning and training. The Workflow4Metabolomics 3.0 e-infrastructure thus not only offers a unique online environment for analysis of data from the main metabolomics technologies, but it is also the first reference repository for metabolomics workflows.

Feature Selection Methods for Early Predictive Biomarker Discovery Using Untargeted Metabolomic Data

Untargeted metabolomics is a powerful phenotyping tool for better understanding biological mechanisms involved in human pathology development and identifying early predictive biomarkers. This approach, based on multiple analytical platforms, such as mass spectrometry (MS), chemometrics and bioinformatics, generates massive and complex data that need appropriate analyses to extract the biologically meaningful information. Despite various tools available, it is still a challenge to handle such large and noisy datasets with limited number of individuals without risking overfitting. Moreover, when the objective is focused on the identification of early predictive markers of clinical outcome, few years before occurrence, it becomes essential to use the appropriate algorithms and workflow to be able to discover subtle effects among this large amount of data. In this context, this work consists in studying a workflow describing the general feature selection process, using knowledge discovery and data mining methodologies to propose advanced solutions for predictive biomarker discovery. The strategy was focused on evaluating a combination of numeric-symbolic approaches for feature selection with the objective of obtaining the best combination of metabolites producing an effective and accurate predictive model. Relying first on numerical approaches, and especially on machine learning methods (SVM-RFE, RF, RF-RFE) and on univariate statistical analyses (ANOVA), a comparative study was performed on an original metabolomic dataset and reduced subsets. As resampling method, LOOCV was applied to minimize the risk of overfitting. The best k-features obtained with different scores of importance from the combination of these different approaches were compared and allowed determining the variable stabilities using Formal Concept Analysis. The results revealed the interest of RF-Gini combined with ANOVA for feature selection as these two complementary methods allowed selecting the 48 best candidates for prediction. Using linear logistic regression on this reduced dataset enabled us to obtain the best performances in terms of prediction accuracy and number of false positive with a model including 5 top variables. Therefore, these results highlighted the interest of feature selection methods and the importance of working on reduced datasets for the identification of predictive biomarkers issued from untargeted metabolomics data.

Systems Metabolomics for Prediction of Metabolic Syndrome

The evolution of human health is a continuum of transitions, involving multifaceted processes at multiple levels, and there is an urgent need for integrative biomarkers that can characterize and predict progression toward disease development. The objective of this work was to perform a systems metabolomics approach to predict metabolic syndrome (MetS) development. A case-control design was used within the French occupational GAZEL cohort (n = 112 males: discovery study; n = 94: replication/validation study). Our integrative strategy was to combine untargeted metabolomics with clinical, sociodemographic, and food habit parameters to describe early phenotypes and build multidimensional predictive models. Different models were built from the discriminant variables, and prediction performances were optimized either when reducing the number of metabolites used or when keeping the associated signature. We illustrated that a selected reduced metabolic profile was able to reveal subtle phenotypic differences 5 years before MetS occurrence. Moreover, resulting metabolomic markers, when combined with clinical characteristics, allowed improving the disease development prediction. The validation study showed that this predictive performance was specific to the MetS component. This work also demonstrates the interest of such an approach to discover subphenotypes that will need further characterization to be able to shift to molecular reclassification and targeting of MetS.

Weight for gestational age and metabolically healthy obesity in adults from the Haguenau cohort

Background An obesity subphenotype, named ‘metabolically healthy obese’ (MHO) has been recently defined to characterise a subgroup of obese individuals with less risk for cardiometabolic abnormalities. To date no data are available on participants born with small weight for gestational age (SGA) and the risk of metabolically unhealthy obesity (MUHO). Objective Assess the risk of MUHO in SGA versus appropriate for gestational age (AGA) adult participants. Methods 129 young obese individuals (body mass index ≥30 kg/m²) from data of an 8-year follow-up Haguenau cohort (France), were identified out of 1308 participants and were divided into 2 groups: SGA (n=72) and AGA (n=57). Metabolic characteristics were analysed and compared using unpaired t-test. The HOMA-IR index was determined for the population and divided into quartiles. Obese participants within the first 3 quartiles were considered as MHO and those in the fourth quartile as MUHO. Relative risks (RRs) and 95% CI for being MUHO in SGA versus AGA participants were computed. Results The SGA-obese group had a higher risk of MUHO versus the AGA-obese group: RR=1.27 (95% CI 1.10 to 1.6) independently of age and sex. Conclusions In case of obesity, SGA might confer a higher risk of MUHO compared with AGA.

Dietary intake in young adults born small or appropriate for gestational age: data from the Haguenau cohort

Objectives Compare the dietary intake of young adults born small for gestational age (SGA) versus those born appropriate for gestational age (AGA). Design Cross-sectional analysis. Setting Data at the 8-year follow-up Haguenau cohort (France). Data from 229 AGA-born adults and 172 SGA-born adults with available dietary information are presented. Methods Dietary intake was based on a food questionnaire including 19 items. The χ2 test was run to compare intake between SGA-born and AGA-born individuals. An a priori score was calculated based on the adherence to recommendations from the French Nutrition and Health Program and included 8 components with the lowest value indicating a lower adherence to recommendations. The score was then divided into quartiles. Relative risks and 95% CIs, controlling for age and sex in multivariate analysis, were calculated in order to determine the risk of belonging to the first versus the second to the fourth quartiles in SGA-born and AGA-born individuals. Results Intakes of SGA-born adults indicated that they consumed more meat, sugar and less fish than AGA-born individuals (all p<0.05). Multivariate analyses with adjustment for age and sex showed that the relative risk of belonging to the first quartile versus the other three quartiles did not disclose any significant difference in SGA-born versus AGA-born participants: RR=0.92 (95% CI 0.65 to 1.30). Conclusions Aside from the differences found by univariate analyses, no significant differences were obtained in multivariate analyses. Findings suggest that parameters of fetal programming are more associated with the development of metabolic syndrome in adulthood rather than dietary patterns.