Many interesting biological processes cannot be modelled by molecular simulations because they take place in time scales inaccessible for current computers. This can be addressed either by parallelisation, i.e. splitting the task to be calculated on a large number of computers, or by application of enhanced sampling techniques. Here we report a new approach combining both strategies. Moving hills method inspired by metadynamics [1] simulates a series of replicas of the studied system in parallel with a shared history-independent bias potential. The bias potential depends on the distribution of selected low-dimensional degrees of freedom (collective variables) among all replicas. The method was tested on model energy profiles, alanine dipeptide (Ace-Ala-Nme) in water and vacuum, cis/trans isomerisation of Ace-(Pro)n-Nme and other molecular systems.
The work was supported by Czech Science Foundation (15-17269S).