[gmx-users] Single long simulation versus multiple short

Tom Dupree t.dupree at unsw.edu.au
Wed Jul 27 01:23:22 CEST 2011

Greetings all,
I am quite interested in this discussion, and wondered if some people would like to add how they would assess the length their MD simulations. I am currently simulating HIV-1 RT for 1 ns and seem to have very flat energy profiles for almost anything (energy wise) I care to measure and yet ligand position/conformation continues to change throughout the simulation. 


Widya Desmarani wrote:
> Dear gromacs user,
> I have been trying to look for an answer for my following question from 
> our forum but still couldn't manage to find one. Probably it is trivial 
> but I am not sure.
> Instead of running a single relatively long simulation (say for about 30 
> ns), is it acceptable if we simulate multiple short simulations (say 15 
> simulations where each of them is 2 ns), and then, all the resulted 
> trajectories are merged/concatenated into one single trajectory with the 
> amount of total simulation 30 ns? I am interested in investigating bulk 
> properties of liquid, such as compressibility, diffusion, and dielectric. 

There is no need to concatenate the trajectories, and moreover it may give you a 
misleading result.  Every simulation needs time to equilibrate.  In the limit of 
infinite sampling, all simulations should converge to the same properties, on 
average.  Whether or not your simulation can produce stable properties in 2 ns 
is something you'll have to decide.  If 2 ns is insufficient, then you're still 
not collecting data that are as reliable as you would get from a longer 

Multiple simulations from independent starting velocities and/or configurations 
are generally expected.  A single simulation does not necessarily tell you the 
whole truth.



Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list