By Richard Norman
This conference brought together researchers in the field of biomolecular simulations to discuss recent advances, challenges, and applications. Some of the main discussion points included:
- Free energy calculations: There was a focus on improving the accuracy of free energy calculations, which are crucial for drug discovery. New methods and tools were presented, such as openforcefield.org and Spice dataset for training ML potentials.
- Coarse-grained models: Coarse-grained models offer a way to simulate larger systems and longer timescales compared to atomistic models. The conference featured discussions on the development and application of coarse-grained models, including Martini for simulating whole cells and ML-based coarse-grained models.
- Integrative modeling: Combining data from different experimental techniques (e.g., cryo-EM, SAXS) with simulations is becoming increasingly important. The conference highlighted progress in this area, showcasing tools like HADDOCK3 and Gapstop.
- Force fields: The accuracy of simulations relies on the quality of force fields. Discussions included the limitations of current force fields and efforts to improve them, such as prosECCo75 and GRAPPA (equivariant NN potentials).
- Specific applications: The conference covered a wide range of applications of biomolecular simulations, including protein design, drug discovery, RNA-targeted therapeutics, and studying diseases like Alzheimer’s.
Additionally, some recurring themes emerged:
- Importance of collaboration: The conference emphasized the value of collaboration between researchers from different disciplines (e.g., computational scientists, experimentalists) to advance the field.
- Challenges and limitations: While significant progress has been made, researchers acknowledged challenges like the high computational cost of simulations and limitations of current force fields.
- Future directions: The talks highlighted promising future directions, such as using machine learning to improve force fields and develop new coarse-grained models.
