DESCRIPTION :
The Doctoral Nexus proposed by the ExposUM Institute are networks of 3 to 4 PhD students from different disciplines and affiliated to at least two different research units. Compared with a traditional PhD, taking part in a Doctoral Nexus will encourage the ability to work in a team and to design projects in a transdisciplinary way while deepening one's own field of expertise. A specific teaching programme will be offered and the doctoral students concerned will also have the opportunity to organise a seminar within the Nexus network. Theses are funded from the outset for 4 years, including the PhD student's salary and an environmental allowance.
Context of the PhD Thesis:
Rheumatoid arthritis (RA) is a complex autoimmune disease triggered by the interaction of genetic, environmental, and personal factors. The PROMESS project is the first national French cohort of individuals at risk of RA, featuring comprehensive biological, clinical, and exposomic phenotyping. It offers a unique opportunity to explore the mechanisms underlying the transition to disease and to develop predictive models that integrate both objective exposures and their subjective perception.
Supervision: The PhD will be jointly supervised by Sandra BRINGAY (Professor, ADVANSE team, LIRMM), Lylia Abrouk (Associate Professor with HDR, on delegation at MISTEA), and Zübeyir Salis (Junior Professor, INSERM Chair).
Zübeyir Salis will be involved across all phases of the project, bringing expertise in semantic modeling, dynamic ontologies, and machine learning, while playing a key role in the understanding and use of data from the PROMESS cohort. As co-leader of PROMESS data analysis within the Nexus framework, he will contribute directly to data structuring, integration into the hybrid model, and interpretation of results from an interdisciplinary and applied perspective.
Laboratories: LIRMM / PhyMedExp
Expected Results:
* A dynamic ontology dedicated to RA and the exposome
* A hybrid predictive model that is customizable, interpretable, and reusable
* A methodological contribution to AI-ontology integration in environmental health research
This project is part of a Nexus involving the groups of IMMEDIATO DAIEN Claire, BOURINET Emmanuel, Rachel AUDO, Emilie OLIE, Marie CHANCEL and Laurent CHICHE
- main mission: This thesis aims to develop a hybrid predictive model for rheumatoid arthritis (RA) by integrating semantic resources in the form of evolving ontologies with machine learning approaches. The central idea is to enrich machine learning models with ontologies that represent both RA and the exposomic factors associated with the disease. The model will enable the integration and analysis of two complementary types of data:
* Objective data: environmental measurements (pollutant levels, living conditions, occupational exposure, etc.).
* Subjective data: individual perceptions, lifestyle habits, and personal experiences.
In particular, this hybrid model will facilitate a more thorough exploration of the complex relationships between exposures and diseases, thereby improving prediction accuracy compared to traditional approaches. It will also follow an evolutionary approach, allowing the ontology to be continuously enriched through the integration of new data and knowledge. This synergy between dynamic ontology and machine learning aims not only to refine predictive capabilities but also to enhance data exploration and interpretability in the context of RA.
The objectives of this thesis are:
1. To build an evolving ontology of RA and its associated exposome.
2. To integrate the knowledge from this ontology into supervised and unsupervised machine learning models.
3. To connect objective data (pollutants, biomarkers) with subjective data (questionnaires, personal experience of exposure).
4. To apply and validate the model using data from the PROMESS cohort.
- activities: Methods and Timeline
Phase 1 - Pre-PROMESS (Year 1 to early Year 2):
* Development of the exposome-RA ontology, covering biological, environmental, and psychological factors.
* Structuring of vocabularies and ensuring interoperability with other semantic resources.
* Training of initial models on test datasets outside of the PROMESS cohort.
The ontology will model RA risk factors, taking into account biological, environmental, behavioral, and socio-economic dimensions. It must be evolving, allowing for the dynamic integration of new findings from medical and environmental research. Once built, the ontology will be used to populate and enrich RA-related databases.
Phase 2 - PROMESS Application (Year 2-3):
* Population of the ontology with data from the PROMESS cohort (exposome, patient-reported outcomes [PROs], mucosal biomarkers).
* Semantic enrichment of machine learning models (feature engineering, ontological constraints).
* Application of the model to predict the clinical transition to RA within two years.
Phase 3 - Exploitation (Year 3-4):
* Continuous improvement of the model with newly acquired PROMESS data.
* Evaluation of model transferability to other potential cohorts.
* Scientific dissemination, including publications and the release of a FAIR-compliant ontology.
Code d'emploi : Ingénieur en Bioinformatique (h/f)
Domaine professionnel actuel : Biologistes
Niveau de formation : Bac+5
Temps partiel / Temps plein : Plein temps
Type de contrat : Contrat à durée déterminée (CDD)
Compétences : Intelligence Artificielle, Analyse des Données, Bases de Données, Machine Learning, Nexus 1000, Feature Engineering, Esprit d'Équipe, Enseignement, Biologie, Amélioration des Processus d'Affaires, Travaux Cliniques, Laboratoire d'Analyses Médicales (LAM), Conceptualisation, Personnalisation, Écologie, Économie, Santé Environnementale, Psychologie, Modélisation Prédictive, Analyses Prédictives, Polyarthrite Rhumatoïde, Facteur de Risque, Sémantique, Vocabulaires
Courriel :
c-daien@chu-montpellier.fr
exposum-aap@umontpellier.fr
lylia.abrouk@u-bourgogne.fr
sandra.bringay@lirmm.fr
zubeyir.salis@inserm.fr
Type d'annonceur : Employeur direct