On Fri, 2014-04-25 at 08:53 +0200, Soumendranath Bhakat wrote:
> Hi Arunima;
>
> Just a quick answer. Always whenever you are starting a production run from
> equil.rst please put irest=0 in the md.in file and you can run 4
> independent MD production runs from that equil.rst point with different
> initial velocities.
>
> Say for example you run 5ns 4 times production run putting irest=0 from
> equil.rst you will get 4 trajectories. Now in the input file put suppose
> starting frame=1; end frame=5000; interval=5 that means you are processing
> 1000 frames for all 4 trajectories. Now you will get 4 different mmpbsa
> outputs taking in account 4 different trajectories. The best way is to take
> average of all of them. This practice is good in multiple MD simulation as
> it removes statistical, solvatiojn, force field related errors.
There are a couple things I would change about this. First, MMPBSA.py
can read as many trajectories as you want and will average over all
snapshots, so there's no need to do this in 4 steps. The more
complicated a workflow is, the easier it is to make a mistake.
Second, simply setting irest=0 is generally a bad idea. In this case,
velocities are assigned from a MB distribution to "tempi", which has a
default value of 0 (so if irest=0, you also need to set tempi to the
same value as temp0). That said, if you use a stochastic thermostat
(ntt=2 or ntt=3), there is really no benefit to irest=0 as the
velocities decorrelate rapidly.
I believe some of the studies you are referring to would start 5 to 10
short simulations from the _same_ starting structure with different
velocity distributions.
All the best,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Apr 25 2014 - 04:30:04 PDT
