Enhanced Sampling Simulations of RNA-peptide Binding using Deep Learning Collective Variables

Enhanced sampling (ES) simulations of biomolecular recognition such as binding of small molecules to proteins and nucleic acids targets, protein-protein association, and protein-nucleic acids interactions have been gaining significant attention in the simulation community due to their ability to sample long timescale processes. However, a key challenge in implementing collective variable (CV)-based enhanced sampling methods is the selection of appropriate CVs that can distinguish the systems metastable states and, when biased, can effectively sample these states. This challenge is particularly acute when simulating the binding of a flexible molecule to a conformationally rich host molecule, such as the binding of a peptide to an RNA. In such cases, a large number of CVs are required to capture the conformations of both the host and the guest, as well as the binding process. In our work, we employed the recently developed Deep Targeted Discrimination Analysis (DeepTDA) method to design CVs for the study of the binding of a cyclic peptide, L22 to a TAR RNA of HIV as a prototypical system. These CVs were used in the on-the-fly probability-based enhanced sampling and well-tempered metadynamics simulations to sample reversible binding and unbinding of L22 peptide to the TAR RNA target. The enhanced sampling simulations revealed multiple binding and unbinding events, which enabled the calculation of the free energy surface for the peptide binding process. Our results demonstrate the potential of the DeepTDA method for designing CVs to study complex biomolecular recognition processes.