Berlin 2024 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
BP: Fachverband Biologische Physik
BP 21: Poster IIIb
BP 21.8: Poster
Mittwoch, 20. März 2024, 11:00–14:30, Poster C
Developing coarse graining RNA force fields via Machine Learning — •Anton Dorn1 and Alexander Schug1,2 — 1Jülich Supercomputing Centre, Jülich, Germany — 2Steinbruch Centre for Computing, Karlsruhe, Germany
In Protein structure prediction there have been massive improvements recently due to deep learning driven exploration of the rich experimental data. A direct transfer, however, of these methods to RNA structure prediction is impossible due to much sparser experimental data for RNA. Still, the combination of molecular force fields with constraints derived from statistical analysis of genomic data such as direct coupling analysis can lead to good quality structure predictions also for RNA. Here, we want to optimize the accuracy of the employed coarse-grained RNA force field for the molecular simulations by employing machine learning techniques. The data sparsity can here be alleviated by building on established atomistic RNA force fields. In a first step we show the viability of this approach by focusing on small RNA molecules in Molecular Dynamics simulations. We explore different bead numbers for the coarse graining to determine the best approximation.
Keywords: Molecular Dynamics; RNA; Machine Learning; Coarse Graining