I did some 2700 ns MD simulation of two protein ligand complexes in Gromacs. However I kept only 0.5 nm distance between protein-surface and box edge. Is it acceptable?
You probably have a lot of minimum image violations, given that most force fields require at least 1.0 nm for most short-range cutoffs. The rule of thumb is to use that short-range cutoff as the minimum solute-box distance to avoid image violations and spurious water ordering effects. Reducing the box size to speed things up has its price...
Probably NO!.. I am not getting image violations during dynamics.
What were the cutoffs used during the run? If you're not getting image violations, then your cutoffs are too short, which is an even bigger problem.
These are the cut-off values:
rlist = 1.0
rcoulomb = 1.0
rvdw = 1.0
vdw_type = cut-off
#Richard,
I performed the MD simulation 100ns time period of 27 proteins. And in all systems the the distance between protein surface and box edge were kept 0.5/6 nm. The proteins are tri-domain in nature with 300-320 AA length.
I must disagree with what Richard has said. The protein's size has no bearing on the acceptable solute-box distance. With a 0.5-nm distance, a small protein will have a 1.0-nm image distance, and so too will a large protein. As you can see, the size of the protein has no bearing on this requirement. The required minimum distance is a function of the nonbonded settings used for the simulation. Now, in the case of a disordered protein or an unfolding simulation, of course a larger cutoff is needed, but the minimum image there still obeys the same requirements, only the calculation is done based on the unfolded state, not the folded state.
I agree with Justin Lemkul , in case of globular, ordered proteins like the ones I'm running right now, I will not go for a 1.0 nm cutoff provided I have adequate computational facilities. Understand the things, "cutoff"s are really the limits for nb calculations etc (coulombic, vdw etc.). So, even when you are going to perform some kind of unfolding-folding simulations to explore the conformational space, you have to give a bigger cutoff. And also, you must avoid image overlap.
As far as I can remember, there's a paper published in the early days, probably in 1992 which discussed about how cutoff values affect the stability of complexes in simulation. If you really want to speed up your simulation, you have a hell lot of options.
- Badges
- Science topic