DynaRepo: from static frozen state to the motion, where function truly lies
Biological molecules such as proteins, RNA, and DNA, are the engines of life, forming complexes that orchestrate everything from gene expression to immune defense. Their 3D structures are critical, but their dynamic movements, influenced by environmental factors like temperature or molecular binding, are equally essential. These motions can enhance function or lead to malfunctions, as seen in diseases like cancer or neurodegenerative disorders. In particular, in a recent study we highlighted the importance of identifying communication pathways within macromolecular complexes, which play a critical role in mediating dynamic interactions and functional outcomes [1]. While tools like AlphaFold have revolutionized static structure prediction, they overlook this dynamic aspect, leaving a significant gap in understanding molecular behavior. DynaRepo, powered by 300,000 GPU hours on Jean Zay, tackles this challenge by generating a vast dataset of molecular dynamics (MD) simulations, offering a new lens to explore life’s molecular dance.
We curated high-quality structures from databases such as PDBbind (for protein-protein complexes), SAbDab (for antigens), and benchmarks (for transient proteins and nucleosomes), resulting in simulations of 450 complexes and 270 single proteins. Each was simulated in three replicates of 500 nanoseconds, yielding over 1.1 milliseconds of data. Unlike traditional repositories focusing on isolated molecules, DynaRepo emphasizes macromolecular complexes, including protein-nucleic acid assemblies, where flexibility plays a pivotal role. This focus is unique, providing a rich resource for studying dynamic interactions that drive biological functions.
The first MDDB node in France for FAIR sharing of MD data
As France’s first node in the European MDDB initiative, DynaRepo is hosted at Inria and integrates with a federated network to share data globally [2]. The dataset includes trajectories, topologies, and pre-computed analyses like RMSD (root-mean-square deviation) and clustering, all accessible at https://dynarepo.inria.fr/. Adhering to FAIR principles (Findable, Accessible via APIs, Interoperable with standard formats, and Reusable with open protocols), it invites contributions from researchers worldwide. Building on more than 1.1 milliseconds already processed for 700 biological systems, we aim to expand to additional complexes, with a focus on protein-nucleic acid interactions critical for gene regulation.
Toward dynamic-aware AI models for macromolecular systems
Future plans include expanding the dataset with protein-nucleic acid complexes, a field crucial for exploration given its role in genetic regulation. Collaborations with national and international labs, supported by Jean Zay’s H100 GPUs, ensure scalability and integration with Europe’s research infrastructure. The project also leverages MDDB’s framework for long-term sustainability, making DynaRepo a cornerstone for advancing molecular science. DynaRepo isn’t just a database, it’s a paradigm shift. By illuminating molecular dynamics, it empowers AI to predict not only shapes but also behaviors, revolutionizing our approach to health, innovation, and disease. As computing power grows and datasets expand, this resource will continue to evolve, bridging the gap between structure and function with unprecedented clarity.
[1] Bheemireddy, S., González-Alemán, R., Bignon, E., & Karami, Y. (2025). Communication pathway analysis within protein-nucleic acid complexes. J. Chem. Theory Comput. 2025, 21, 17, 8255–8266.
[2] Mokhtari, O., Bignon, E., Khakzad, H., & Karami, Y. DynaRepo: The repository of macromolecular conformational dynamics. bioRxiv, 2025.
[3] Mokhtari, O., Grudinin, S., Karami, Y., & Khakzad, H. DynamicGT: a dynamic-aware geometric transformer model to predict protein binding interfaces in flexible and disordered regions. bioRxiv, 2025.
A key number :
Total simulation time: more than 1 100 microsecondes of molecular dynamics simulations for more than 700 macromolecular systems.
Definitions:
1. Molecular Dynamics (MD) : A computational method that simulates the movements of atoms and molecules over time, revealing the flexibility and conformational changes essential to biological functions.
2. Macromolecular Complexes : Assemblies of proteins, RNA, and DNA that perform key cellular processes, such as gene regulation; their dynamics influence interactions and disease.
3. FAIR Principles : Standards for Findable, Accessible, Interoperable, and Reusable Data, promoting open sharing in science to accelerate discovery.