(525b) Sampling the Conformational Space of Proteins Using Anisotropic Network Model Modes Driven Metadynamics

Complex biological macromolecules such as proteins carry within their functional, native structures the key to many essential life processes. While molecular simulations are used to explore protein conformational changes at the atomistic scale, they are often limited by accessible temporal scales. To overcome this limitation, enhanced sampling methods have been used. Parallel-tempering metadynamics (PTMetaD) is an approach that facilitates the efficient sampling of the free energy landscape based on the biasing of a set of collective variables and high temperatures that accelerate dynamics. A significant bottleneck in the success of PTMetaD is the its proneness to be affected from the choice of collective variables (CVs). In this work, we propose a new framework to rationally design efficient CVs for PTMetaD that minimize large hysteresis effects. The algorithm involves using the slow modes calculated from α-carbon-based anisotropic network model (ANM) as CVs to rationally drive the system along the highest collective modes of motion. By only using the intrinsic dynamic modes of motion that are accessible to the protein to bias simulations, convergence of the free energy landscape improves significantly compared to PTMetaD based on conventional CVs. Through simulations of multiple proteins that exhibit a diverse set of biological functions and conformational changes, we show how CVs derived from ANM modes can facilitate efficient and reliable characterization of conformational changes.