Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 15: Computational Biophysics III

BP 15.6: Vortrag

Mittwoch, 11. März 2026, 11:15–11:30, BAR/SCHÖ

From Data to Discovery: Machine Learning Force Fields for Fast and Accurate Ligand-Protein Screening — •Sergio Suárez-Dou¹, Miguel Gallegos¹, Hamza Ibrahim², Joshua T. Berryman¹, Andrea Volkamer², and Alexandre Tkatchenko¹ — ¹Department of Physics and Materials Science, University of Luxembourg, Luxembourg — ²Data Driven Drug Design, Center for Bioinformatics, Saarland University, Germany

Pretrained Machine Learning Force Fields (MLFF) are transforming computational chemistry by combining speed and accuracy. In drug discovery, predicting binding energies and affinities is key to identifying viable candidates. Among MLFFs, SO3LR excels in both performance and efficiency, enabling accurate ligand binding predictions at low computational cost, surpassing semiempirical methods.

However, data coverage remains a challenge. While datasets like OMol25 and QCell offer broad molecular diversity, drug-like chemical space requires more targeted representation. To address this, we are developing a dataset optimised for docking evaluations, covering key regions of drug-like space. Our goal is a pretrained model capable of predicting drug-ligand affinities with PBE0+MBD DFT-level accuracy, bridging quantum precision with high-throughput screening and accelerating drug discovery.

Keywords: Machine Learning Force Fields (MLFF); Drug discovery; Binding Affinity; Drug-like Chemical Space