A short MD simulation is often criticized by a referee, however, depending on the phenomenon we observe like in my case side chain mobility, what would be the best simulation time in explicit water for my system?
I am supposed to evaluate the sampling time by looking at the convergence of a reaction coordinate. If this convergence is reached at 2ns, then is 10ns of simulation time ok?
If the sampling time is enough for the phenomenon you want to observe, then a referee should not be complaining about it. However, a single simulation will yield a single outcome. Think of an experiment: you never measured whatever magnitude only once, but look for statistical values. I would suggest to sample the same event, but starting with different initial velocities on each atom, or with different atomic position (i.e., different initial conformations). This way you can tell that you have a set of independent simulations to test their convergence towards the corresponding expected result.
Dear Laurent,
in reality it's impossible to say that setten time of MD simulation is long enough. Conformational space, in general, is so complex that even if we have pretty good knowledge about system - it's hard to predict sufficent time of simulation.
You didn't tell much about your system. But you can always try to check the MD convergence using specific tricks, like games with PCA or calculating the RMSIP.
Side chains undergo conformational changes, if your simulation time can allow the sampling of all possible conformations till some threshold (may be 90% ) or some other lower limit which you feel is enough. Then you can convince the reviewer that you have allowed the simulation to mimic real system's behaviour (provided the force field is apt for the system). This way you can evaluate the simulation time needed.
You can use "bin" system to collect the sampling stats. But, once you reached a threshold (achieved relatively faster) it really needs a very long simulation time to sample that extra 10% of space. So do not try to sample all space.
Other researchers in the field perform multiple simulations with different random number seeds and/or starting structures (e.g. taken out of the minimization phase) to check for consistency of outcome).
You can perform replica exchange simulations (that uses different temperatures) in order to collect a wider ensemble of conformations. Or simply perform an MD at an elevated temperature to promote the motions (be careful do not unfold the protein). You can evaluate thereafter with your MD how long simulation is needed to observe the various side-chain conformations. This is a first glimpse to the problem you are adressing.
- Badges
- Science topic
- Similar topics
- Chemistry
- Theoretical Chemistry
- Computational Chemistry
- Molecular Dynamics