Free-Energy Surface Prediction by Flying Gaussian Method

Molecular simulations are computationally expensive, especially in systems with multiple free energy minima. To address this problem many enhanced sampling techniques have been developed. Metadynamics uses a bias potential defined as a sum of Gaussian hills in space of few (one or two) collective variables. This bias potential disfavors states that have been visited since the beginning of the simulation. Multiple walker metadynamics simulates the system in multiple parallel replicas. The bias potential disfavors states that have been visited since the beginning of the simulation in any replica. Flying Gaussian method presented here also simulates the system in multiple parallel replicas. It disfavors states that are, at certain moment, similar in two or more replicas. It was demonstrated on Alanine Dipeptide in vacuum and water, cis/trans-isomerisation of Proline-containing peptides and Met-enkephalin.