AMBER Archive (2009)
Subject: Re: Re: [AMBER] MD simulation with a partially flexible system
From: Carlos Simmerling (carlos.simmerling_at_gmail.com)
Date: Wed Feb 18 2009 - 06:35:40 CST
the reduction in cost only ahppens if you can skip the calculation of
forces for those atoms and the associated distances. for PBC, that can
be done, but constant P isn't correct and the reciprocal space
calculation isn't faster, only the direct. for GB, you still need to
calculate effective radii of fixed atoms, since they can change if
atoms near them move and these radii are needed for the forces on the
moving atoms. on the whole it is difficult to speed this up
efficiently, though I do recall reading a paper about it. In Amber
there can be speedup with belly but there are restrictions on which
atoms can be frozen. I'd suggest giving it a try and seeing how it
goes- but I would not expect huge speedups in most cases. another
approach might be to consider not including most of the other atoms,
only a buffer region of restrained or fixed atoms at the end of your
moving atoms zone.
On Tue, Feb 17, 2009 at 9:19 PM, Jeffrey <jeffry20072008_at_yahoo.cn> wrote:
> Hi Carlos,
> Thanks for the reply. What I want to do is trying to reduce the computational cost by focusing only a small part of the whole system. Since we only concerns the dynamical change of a subdomain, it can be reasonable to allow only the subdomain and its neighbouring atoms to move while keeping the rest atoms rigid during the simulation, which I thought the computational cost would reduce greatly compared to keeping the whole system flexible.
> Thanks very much for the time.
>>can you be more clear about how you want to reduce the computational
>>expense? are you following an example from the literature, or just
>>trying it out to see if it can be done?
>>you could use the belly option, but the details of how well this works
>>depend on whether your simulation is periodic, or if you use GB
>>solvent, etc. you won't always get much improvement in speed, and also
>>the belly option can have issues with PBC, you can find out more about
>>that in the archives.
>>On Tue, Feb 17, 2009 at 9:38 AM, Jeffrey <jeffry20072008_at_yahoo.cn> wrote:
>>> Dear amber users,
>>> I'd like to perform a microsecond-scale MD simulation to see the change of a subdomain with time. So it is necessary to reduce the computational expense by keeping only the subdomain and its surrounding atoms to move while keeping the rest of the system as rigid body during the course of simulation. I checked the amber mannual and didn't find a good choice.
>>> Does anyone know whether this job can be done in AMBER package?
>>> Thanks very much.
>>> AMBER mailing list
>>AMBER mailing list
> AMBER mailing list
AMBER mailing list