(169ax) Using Deep Learning to Accelerate the Molecular Simulations and Predict the Kinetics of RNA Folding

Understanding the process of RNA stem-loop folding and unfolding is crucial in biology. The poster will address the computational challenge of simulating RNA stem-loop folding due to the complex folding landscape that requires extensive computation. The research explores how to study the folding of RNA stem-loops using a deep learning-driven simulation method. We adapted DeepDriveMD (DDMD), a deep learning technique, to simulate RNA stem-loop folding dynamics. DDMD adaptively learns a low-dimensional latent representation from an ensemble of running MD simulations. DDMD then guides the simulations toward the undersampled regions while optimizing the resources to explore the relevant parts of the phase space. The method achieves reasonable free energy landscape prediction at room temperature and identifies relevant slow degrees of freedom in RNA folding. The kinetic rates of transition between the stable and metastable states were calculated using Markov state models. The qualitative analysis of the latent space indicated that the computational framework could capture the distinct conformations within the RNA folding process. DDMD can simulate RNA folding more efficiently than conventional methods, providing insights into phase space and system kinetics without extensive computational costs. This provides a feasible and more rapid approach for exploring the complex process of RNA folding, with broad implications for computational biology.