Thèse Modélisation Informée par Données Expérimentales de la Communication Allostérique dans Systèmes Biomoléculaires H/F

Université Paris Cité

Établissement : Université Paris Cité
École doctorale : Chimie Physique & Chimie Analytique de Paris-Centre
Laboratoire de recherche : Laboratoire de Biochimie Théorique
Direction de la thèse : Jérôme HÉNIN ORCID 0000000325404098
Début de la thèse : 2026-10-01
Date limite de candidature : 2026-05-11T23:59:59

La régulation allostérique permet aux protéines de transmettre des signaux entre sites fonctionnels distants via leurs fluctuations conformationnelles. Ce projet vise à développer un cadre computationnel informé par des données expérimentales pour reconstruire les graphes de communication qui décrivent comment ces signaux se propagent à travers la structure protéique. À partir d'ensembles conformationnels issus de simulations de dynamique moléculaire et de méthodes basées sur l'IA, des graphes de communication seront construits puis raffinés en intégrant des données expérimentales telles que cinétique enzymatique, mutagenèse à haut débit, et mesures de dynamique conformationnelle (RMN, HDX), dans un cadre bayésien. Cette approche permettra d'identifier les voies dominantes de communication et de comprendre comment les mutations ou perturbations les modifient. La méthode sera appliquée à des systèmes biomédicalement pertinents, notamment KRAS et des phosphatases, afin d'élucider les mécanismes moléculaires des variants pathologiques. Le projet fournira un cadre général reliant dynamique conformationnelle et fonction biologique.

Allosteric regulation enables proteins to control function through communication between distant functional sites mediated by conformational dynamics. Molecular dynamics simulations provide atomistic access to these fluctuations and enable construction of communication graphs describing how structural changes propagate through the protein. However, networks derived solely from simulations may include couplings that are physically plausible but not functionally relevant. Experimental techniques such as deep mutational scanning, kinetics, NMR, and HDX provide complementary information about functional sensitivity and conformational dynamics. Integrating these data provides a principled strategy to refine communication graphs so they reflect experimentally observed functional mechanisms. The Laboratoire de Biochimie Théorique has extensive expertise in molecular simulations and network analysis and has developed tools such as MDiGest, providing a strong foundation for this project.

The objective of this project is to develop an experimentally grounded computational framework to infer communication graphs that describe how regulatory (allosteric) dynamics propagates between distant functional regions in proteins and how perturbations reshape these communication pathways. Starting from simulated conformational ensembles, the project will integrate functional measurements (kinetics, deep mutational scanning) and structural dynamics data (NMR, HDX) to refine network representations within a Bayesian inference framework. This will enable identification of dominant communication pathways and determine how gain-of-function, loss-of-function, and disease-associated variants rewire these mechanisms relative to the wild type. The methodology will be applied to KRAS GTPases and protein tyrosine phosphatases, which provide well-characterized model systems with extensive experimental datasets.

The project will develop an experimentally informed communication graph inference framework combining molecular dynamics simulations, graph-based modeling, and Bayesian inference. Conformational ensembles will be analyzed using the MDiGest software to construct communication graphs whose edge weights reflect dynamical couplings inferred from correlated conformational fluctuations, defining a physically grounded prior network. Experimental functional measurements, including kinetics and deep mutational scanning, will be incorporated as likelihood constraints to refine graph structure and coupling strengths, while excluding mutations dominated by trivial structural or catalytic disruption. Structural dynamics measurements such as NMR and HDX will further constrain the network to ensure consistency with experimentally observed conformational fluctuations and state populations. The resulting posterior graph will provide an experimentally informed representation of protein communication. The framework will be applied to KRAS and protein tyrosine phosphatases to compare communication mechanisms across functional and disease-associated variants.