Mots-Clés
bioinformatics
machine-learning
biostatistics
programming
Description
The main objective of the position is to identify and validate early predictors of major cardiometabolic events using baseline multi-omics data from a deeply phenotyped cohort, with a particular focus on gut microbiota metagenomic profiles and long-term longitudinal follow-up. The cohort (METACARDIS) benefits from particularly rich baseline phenotyping, including:
• gut microbiota metagenomic data,
• metabolomic analyses (blood and urine),
• environmental and behavioral data collected through internationally validated food frequency and physical activity questionnaires.
In 2025, a linkage was performed between French participants of the METACARDIS cohort (more than 700 individuals recruited between 2013 and 2015) and the French National Health Data System (SNDS), making it possible to identify the occurrence of major cardiometabolic events among cohort participants, including the onset of diabetes and cardiovascular events, with comprehensive longitudinal follow-up. The post-doctoral researcjer will assess the added value of metagenomic data for cardiometabolic risk prediction, in comparison with and in addition to established clinical and biological risk factors.
The project addresses major methodological challenges in cardiometabolic risk prediction, including the handling of censored and time-to-event data, integration of high-dimensional multi-omics data, assessment of model calibration and discrimination, model interpretability, and rigorous internal and external validation.
Main activities
The postdoctoral researcher will:
• Design and implement statistical and predictive modeling strategies adapted to longitudinal and multi-omics data.
• Develop and apply cardiometabolic risk prediction models, including survival models and machine learning approaches.
• Compare predictive performance of models with and without metagenomic data (added value assessment, calibration, discrimination).
• Select and adapt the most relevant statistical and algorithmic methods according to biological hypotheses and data constraints.
• Integrate and harmonize heterogeneous data (clinical, biological, metagenomic, environmental).
• Implement advanced data visualization methods for data exploration and results reporting.
• Maintain methodological and technological watch on modeling approaches, machine learning methods, and associated computational infrastructures.
• Contribute to the biological and clinical interpretation of results in interaction with consortium members.
• Write and submit scientific articles to peer-reviewed international journals.
• Contribute to the scientific dissemination of the project (presentations, conferences, collaborations).
Required transversal knowledge
The candidate must have strong expertise in statistical analysis and modeling of biomedical data, applied to observational and longitudinal studies in human health.
He/she should have a solid understanding of cardiometabolic risk factors, principles of analytical epidemiology, and challenges related to clinical event prediction (bias, confounding, temporality, model validation).
Operational knowledge of gut microbiota metagenomic data is expected, particularly for:
• biological interpretation of microbial signatures,
• use of taxonomic and functional data as explanatory variables,
• comparison of models including or excluding omics data.
The candidate must master the fundamentals of multivariate statistical methods and machine learning approaches, including variable selection, model validation, and evaluation of predictive performance.
He/she should be comfortable with scientific computing and data analysis tools, and have proven experience in programming languages, particularly R. Experience with Python and/or server-based computing environments is an asset.
A good understanding of health data derived from medico-administrative databases (or, alternatively, of related issues such as coding, censoring, and longitudinal follow-up) is desirable.
Proficiency in English (B2 to C1 level), particularly for scientific writing and communication, is essential.