Thèse Identification de Biomarqueurs et Modélisation Prédictive de la Réponse à l'Immunothérapie Basée sur le Microbiote Intestinal H/F - Université Paris-Saclay GS Life Sciences and Health

Établissement : Université Paris-Saclay GS Life Sciences and Health
École doctorale : Cancérologie : Biologie - Médecine - Santé
Laboratoire de recherche : MIcrobiota and MucOsae for cancer immunoSurveillAnce
Direction de la thèse : Lisa DEROSA ORCID 000000030527296
Début de la thèse : 2026-10-01
Date limite de candidature : 2026-04-07T23:59:59

Les inhibiteurs de points de contrôle immunitaire (immune checkpoint inhibitors, ICI) ont profondément transformé la prise en charge des cancers avancés. Toutefois, une proportion importante de patients présente une résistance primaire ou développe des toxicités immuno-induites, soulignant la nécessité d'identifier les déterminants biologiques de la réponse thérapeutique. Des données croissantes indiquent que des facteurs liés à l'hôte, notamment la composition du microbiote intestinal, le métabolisme, l'état immunitaire systémique et les habitudes alimentaires, jouent un rôle majeur dans la modulation de l'immunité antitumorale. Malgré ces avancées, plusieurs verrous scientifiques persistent : l'absence de biomarqueurs robustes permettant de prédire la réponse aux ICI, la compréhension encore limitée des mécanismes physiopathologiques reliant dysbiose intestinale et résistance thérapeutique, ainsi que le manque d'approches guidées par biomarqueurs pour orienter des interventions ciblant le microbiote. Dans ce contexte, l'unité INSERM U1367 de Gustave Roussy conduit deux essais prospectifs, IMMUNOLIFE et MICADO, visant à constituer une cohorte longitudinale d'environ 1000 patients traités par immunothérapie. Ces études génèrent des données multiomiques intégrées comprenant métagénomique shotgun, métabolomique, immunomonitoring, ainsi que des données nutritionnelles et de mode de vie. Elles permettront d'étudier les conséquences fonctionnelles et physiopathologiques de l'exposition aux antibiotiques et des facteurs alimentaires, notamment avant et après des interventions de modulation du microbiote telles que la transplantation de microbiote fécal allogénique et les interventions nutritionnelles enrichies en fibres. La question centrale est de déterminer si l'intégration de biomarqueurs microbiotiques, métaboliques et immunitaires permet d'identifier les déterminants de la résistance ou de la réponse aux ICI et de guider des stratégies thérapeutiques personnalisées.
Les objectifs scientifiques sont :
1.Valider des biomarqueurs prédictifs associés aux résultats sous immunothérapie à partir de cohortes prospectives et de biobanques rétrospectives, en se concentrant sur quatre axes physiopathologiques majeurs : dysbiose intestinale, dysmétabolisme, immunosuppression, altération de la perméabilité intestinale.
2.Démontrer l'actionnabilité clinique de ces biomarqueurs en évaluant des interventions interceptives ciblant le microbiote.
Dans ce cadre, le projet doctoral repose sur une approche de science des données appliquée à l'oncologie de précision. Le doctorant aura pour mission l'assemblage et l'harmonisation des données multiomiques, l'exploration et le data mining de jeux de données complexes, ainsi que le développement de modèles prédictifs visant à identifier des signatures biologiques d'intérêt clinique. Le travail sera réalisé en interaction étroite avec des immunologistes, biologistes cellulaires, cliniciens et biostatisticiens au sein d'un laboratoire expert en immunologie des cancers.

Over the past decade, cancer immunotherapy has profoundly transformed the management of multiple malignancies previously associated with poor prognosis. Immune checkpoint blockade (ICB) now represents a cornerstone of modern oncology. Despite these advances, a substantial proportion of patients fail to achieve durable clinical benefit. Understanding the biological determinants of treatment response and resistance therefore represents a critical unmet need in oncology.
Accumulating evidence indicates that the gut microbiota is a key regulator of systemic antitumor immunity1,2. Our previous research has demonstrated that a taxonomic microbiome signature, including the TOPOSCORE3, S Score, SIG1 bacterial count, and the relative abundance of Akkermansia spp4-7., is associated with resistance to ICB and poor survival outcomes. These features are emerging as robust biomarkers of an unhealthy gut microbiome. This unhealthy status may arise from cancer-associated inflammation, aging, lifestyle factors such as diet and physical activity, and particularly exposure to concomitant medications. Antibiotic exposure represents one of the strongest iatrogenic drivers of microbiota disruption5,8,9. Antibiotics profoundly reshape microbial ecosystems, increasing SIG1-releated members particularly Enterocloster spp10. Clinical and preclinical studies from our group have consistently demonstrated that this is associated with impaired cancer immunosurveillance and reduced efficacy of ICB. Beyond taxonomic alterations, antibiotics also induce intestinal barrier dysfunction, reflected by perturbations of the mucosal homing axis and circulating markers such as soluble MAdCAM-110, as well as profound functional metabolic disruptions, involving bile acid metabolism (unpublished data).
Advances in next-generation sequencing technologies have enabled identification of microbial fingerprints associated with treatment outcomes. Baseline gut microbiota composition correlates with both response and toxicity to immune checkpoint blockade11. Very preliminary longitudinal analyses indicate that, in responders receiving immunotherapy, the gut microbiota remains relatively stable over time, suggesting ecological resilience of a beneficial microbial state3. In contrast, cytotoxic chemotherapy induces marked microbial perturbations, inflammatory cytokine release, and loss of microbial diversity, potentially impairing immune competence12. These observations indicate that anticancer therapies alone are insufficient to restore a favorable microbiota ecosystem. Consequently, increasing attention has focused on microbiota-centered interventions aimed at reconditioning host immunity. Fecal microbiota transplantation (FMT) has demonstrated the capacity to reshape intestinal ecosystems and restore immunotherapy sensitivity in early clinical studies13,14. Responders exhibit depletion of deleterious SIG1-releated bacterial species linked to treatment resistance. Nevertheless, important limitations hinder the widespread clinical implementation of FMT, including donor variability, regulatory constraints, logistical complexity, and uncertainties regarding long-term safety. These challenges highlight the need for more controllable, scalable, and personalized strategies for microbiota modulation.
Growing evidence supports a diet-microbiota-immune axis as a key determinant of therapeutic response. Diet acts as a major upstream regulator of gut microbial composition and function15. Dietary fiber promotes beneficial fermentative bacteria, increases short-chain fatty acid production, and supports intestinal barrier integrity and immune homeostasis16. In addition, our group demonstrated that a histidine-rich diet was associated with improved outcomes in patients whose gut microbiome lacked dysbiotic pathways converting histidine into the immunosuppressive metabolite imidazole propionate (unpublished data).
We hypothesize that resistance to cancer immunotherapy arises from disruptions of the host-microbiota-immune axis, characterized not only by compositional gut dysbiosis but also by functional metabolic and immune imbalance leading to impaired immune fitness. Integrating metagenomic, metabolomic, and immunological biomarkers will enable the identification of dysbiotic patients, and the implementation of biomarker-guided microbiota-centered interventions aimed at restoring antitumor immunity and improving clinical outcomes.
This framework supports a paradigm shift in oncology, whereby effective cancer immunotherapy may require ecological and metabolic rehabilitation of the host-microbiota system through rational, sustainable interventions compatible with modern oncological treatments.

The project aims to develop an integrated framework to identify, validate, and translate predictive biomarkers of response and resistance to immune checkpoint inhibitors (ICIs) using longitudinal multi-omics data.
Objective 1 : Validation of predictive biomarkers
To validate candidate biomarkers associated with response, resistance and toxicity to ICI across the prospective IMMUNOLIFE and MICADO trials as well as retrospective biobank datasets, focusing on four major pathophysiological drivers :
-gut dysbiosis,
-dysmetabolism,
-immunosuppression,
-gut barrier dysfunction and permeability alterations.
Candidate biomarkers include microbiota-derived scores (e.g., TOPOSCORE), circulating markers of gut-immune interaction (e.g., MAdCAM-1), metabolomic profiles with a focus on bile acids, microbial genes involved in bile acid metabolism, and integrated multi-omics signatures reflecting host-microbiome functional interactions.
Objective 2 : Multi-omics integration and predictive modeling
To develop integrative computational models combining metagenomics, metabolomics, immunomonitoring, and nutritional/lifestyle data in order to identify biological signatures predictive of response, resistance, and toxicity and longitudinal clinical trajectories.
Objective 3 : Demonstration of biomarker actionability
To demonstrate the clinical actionability of validated biomarkers by evaluating their ability to guide microbiota-centered therapeutic interventions within a precision oncology framework, including:
-fecal microbiota transplantation (FMT),
-fiber-based dietary modulation,
-microbiota restoration strategies following antibiotic-associated dysbiosis.
Simulation analyses and predictive modeling outputs generated in Objective 2 will be used to design personalized intervention strategies and define patient stratification algorithms.

The project relies on a data science-driven translational research approach, integrating computational biology with clinical and experimental immunology.
1. Data Assembly and Harmonization
-Integration of longitudinal datasets generated by IMMUNOLIFE and MICADO.
-Harmonization of heterogeneous data types, including clinical variables, treatment exposure, survival endpoints, multi-omics layers, and lifestyle parameters.
-Development of a secure, multimodal analytical database enabling reproducible and scalable analyses.
-Implementation of standardized preprocessing pipelines for metagenomic, metabolomic, and immunological datasets.
2. Exploratory Analysis and Data Mining
-Multi-omics exploratory analyses to characterize global microbial diversity (alpha/beta diversity), taxonomic composition, functional pathways, and metabolite profiles.
-Feature selection strategies adapted to high-dimensional datasets, including regularization approaches and embedded selection within machine-learning algorithms.
-Identification of microbial, metabolic, and immune signatures associated with clinical outcomes using logistic regression, random forest (RF), support vector machines (SVM), and other supervised learning approaches.
-Assessment of interactions between diet, antibiotics exposure, microbiome composition, and immune parameters.
3. Predictive Modeling
-Development of statistical and machine-learning models, including:
oSurvival modeling (Cox regression, penalized survival models),
oSupervised classification models for response/toxicity prediction,
oLongitudinal trajectory modeling to capture intra-individual dynamics over time.
-Use of advanced ensemble methods such as Random Survival Forests (RSF) and dynamic modeling approaches (e.g., DynForest) to integrate time-dependent variables.
-Internal validation through cross-validation and bootstrapping.
-External validation using independent retrospective and prospective datasets.
4. Biological Interpretation and Translational Integration
-Close collaboration with immunologists, clinicians, cell biologists, statisticians, and bioinformaticians (including partners from INRAE, CentraleSupélec, and UNITRENTO).
-Mechanistic interpretation of identified signatures in light of immune pathways, barrier function biology, and microbial metabolism.
-Simulation analyses using models derived from Objective 2 to inform interventional strategies in Objective 3.
-Iterative feedback between computational findings and experimental validation to ensure biological plausibility and translational relevance.

Nous recherchons un(e) doctorant(e) fortement motivé(e), doté(e) d'une formation en science des données, bioinformatique, biostatistique, biologie computationnelle ou discipline quantitative apparentée. Compétences et qualifications requises : Diplôme de Master (ou équivalent) en data science, bioinformatique, mathématiques appliquées, statistiques, informatique, biologie computationnelle ou ingénierie biomédicale. Bonnes compétences en programmation (Python et/ou R). Intérêt pour la recherche biomédicale, la médecine de précision et l'intégration de données multiomiques. Compétences appréciées : Connaissances en données omiques (métagénomique, métabolomique, données cliniques). Connaissances en analyses de survie. Connaissances en manipulation des jeux de données de grande dimension. Qualités personnelles : Aptitude au travail interdisciplinaire. Maîtrise de l'anglais. Autonomie, curiosité scientifique et motivation pour la recherche translationnelle à l'interface entre science des données et immunologie des cancers.