Paris
École Nationale des Ponts et Chaussées, Université Gustave Eiffel, Institut de Biologie Physico-Chimique, École Normale Supérieure and Sorbonne Université, with expertise in mathematics, biophysics and chemistry.
International Research Project
A France–Chicago collaboration
This IRP brings together applied mathematicians, molecular simulation specialists and machine-learning researchers in France and at the University of Chicago.
The project develops mathematically grounded methods for complex molecular systems, with a particular focus on high-performance computing and neural networks. It builds on long-standing collaborations between the French and Chicago teams and supports exchanges across mathematics, biophysics, chemistry and scientific computing.
- Principal Investigators
- Christophe Chipot and Tony Lelièvre
- Partners
- CNRS teams in France and the University of Chicago
- Ambition
- A five-year interdisciplinary research programme
Scientific programme
The project studies the mathematical and computational challenges raised by stochastic molecular dynamics, from equilibrium sampling to rare transitions and reduced models. Machine learning provides new tools, while mathematical analysis is essential to understand their accuracy, robustness and limitations.
- High-dimensional functionsNeural representations for committor functions and other quantities central to molecular modelling.
- Dimension reductionCollective variables, latent representations and effective dynamics for complex systems.
- Generative modelsEfficient sampling strategies combined with statistically exact acceptance and correction mechanisms.
- Machine-learned force fieldsFast, accurate potentials for molecular dynamics, supported by numerical analysis.
- Metastability and rare eventsMathematical models and algorithms for transition pathways, reaction times and transport.
- High-performance computingScalable methods that connect mathematical ideas with production molecular-simulation software.
- Target applicationsLarge collective motions in proteins and protein complexes, biological processes involving chemical reactions, RNA, intrinsically disordered proteins, pharmaceutical targets, biocatalysts and materials.
A three-node network
Nancy
Université de Lorraine, with expertise in molecular modelling, free-energy calculations and biomolecular simulation.
Chicago
University of Chicago researchers in chemistry, molecular engineering, mathematics and statistics.
Kick-off meeting
9–11 September 2026 · Paris
The inaugural meeting at the University of Chicago’s John W. Boyer Center in Paris brought together senior and early-career researchers from the three nodes.
The scientific sessions covered rare-event simulation, committor functions, collective variables, generative sampling, equivariant neural networks and machine-learned interatomic potentials.