I understand metadynamics is an efficient way of sampling by adding bias potential to already-sampled states (conformations etc.). Now I have a protein-ligand system. I wonder if the initial conformation (spatial and geometry) of my system would matter for the efficiency of sampling metadynamics. E.g. there two local minima A and B, with A close to absolute minimum and B far away from absolute minimum on the free energy surface. What difference would there be if I start at A as opposed to B?
Molecular dynamics is a technique which generate a trajectory in the phase space by solving the Newton's equations of motion (or more generally Hamiltonian equations of motion) for some initial conditions (particles coordinates and velocities).
Then, based on the assumption of ergodicity, the time average as obtained by MD trajectory are assumed to be equivalent (the equal) with ensemble averages. The ensemble averages are weighted ensemble averages with Baltzmann factors, where each point in the phase space has a certain probability to be visited. Thus, if we simulate for very long time, then the trajectory will be able to visit all possible points of the phase space, eventually.
Practically, since the fluctuations in MD simulations are of order kBT, not all points are possible to visit, then metadynamics is one of the methods, which is used to enhance the sampling, that is we will be able, possibly, to visit all points in the phase space, independently form the initial conditions. Indeed, since metadynamics adds a bias in each point of trajectory based on its history, eventually, after some times, the energy landscape will be flattered and all points of the phase space will be visited (even the global minimum, but not just the global minimum, all points, of course with different probabilities, that is the global minimum has higher probability or will be visited longer.) This is also the difference of md technique from other optimization techniques, such as Global Optimization Techniques, which stop as soon they find the so called global minimum, which is not the case of molecular dynamics simulations.
Therefore, It should not matter if you start from A or B.
References for metadynamics:
A. Barducci, M. Bonomi, M. Parrinello, Metadynamics, Advanced Review, 1, 826, 2011.
- Badges
- Science topic