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

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 21: Modeling and Simulation of Soft Matter II

CPP 21.4: Vortrag

Dienstag, 18. März 2025, 14:45–15:00, H34

Machine-Learning potentials to understand pairing and stacking at the origin of life — •Laurie Stevens¹, Riccardo Martina², Alberta Ferrarini², and Marialore Sulpizi¹ — ¹Faculty of Physics and Astronomy, Ruhr-Universität Bochum, Germany — ²Chemical Science Department, Università degli Studi di Padova, Italy

When exploring the origin of life, one main question remains open: how did we get from single nucleotides to long RNA and DNA chains which then led to more complex biological structures, following the RNA world hypothesis. More specifically, we are interested in how the nucleotides interactions are able to promote the synthesis of long polynucleotides. Experimental studies suggest that free nucleotides in water spontaneously organize into small molecular columnar phases, promoting the ligation of nucleic acid chains. To uncover the mechanisms behind this self-assembly, we use Molecular Dynamics simulations combined with Machine-Learning approaches.

Ab initio methods are too computationally expensive for the timescale of interest and for the complexity of the investigated systems. To overcome this limitation, we use Neural Network Potentials (NNPs) trained with DeepMD-kit and reinforced by metadynamics. After mastering the static and dynamical properties of a single Adenosine Monophosphate (AMP) in water, we are now investigating the stacking and pairing interactions between several AMPs by predicting the free energy landscape of this system as a function of the relevant degrees of freedom.

Keywords: Machine Learning; Molecular Dynamics; Neural Network Potential; Nucleic Acids; Self-assembly