Re: Computing cost of frozen atoms

From: Marc Q. Ma (qma_at_oak.njit.edu)
Date: Tue Apr 18 2006 - 12:50:59 CDT

Hi,

I guess it is feasible to use VMD and SMD to do this type of
simulations.

1. separate the part you are interested into a different file, and
let NAMD calculate all the forces.

2. From the parts you are interested in, use tcl force to calculate
all the forces you needed from the frozen parts (the electrostatic
contri).

3. Add these tcl forces to the first part.

Thus, this is similar to SMD. I do not have any sample script to do
this directly. But I think the VMD gurus out there will be able to
help you achieve the above -- if you like the idea.

Marc

On Apr 18, 2006, at 12:09 PM, Sterling Paramore wrote:

> The most expensive part of MD is the force routine. Since you'll
> still be calculating the force from the "frozen" atoms, the
> computational cost will be just about the same as it would be
> without frozen atoms.
>
> I suppose that knowing some atoms are frozen could potentially lead
> to a different or more efficient parallelization algorithm. But I
> don't know how NAMD handles it, and I doubt you'll get a
> significant speed up.
>
> -Sterling
>
> Jacob Pøhlsgaard wrote:
>
>> Hi people
>> I'm looking to simulate a very small area of a very large
>> complex.For this I'm going to freeze some residues, restrain
>> others and allow the atoms in the area of interest to move freely.
>> My question is if anyone knows how much computational cost I
>> have to pay by including the whole (~150.000 atoms) complex as
>> frozen atoms? Is there some general rule of thumb?
>> Is it a good idea to include all the residues at all? I'll have
>> atoms way beyond the electrostatic cutoff, but as I understand it,
>> there's still a a contribution to electrostatics that far away
>> since I'm using the PME method.
>> Best regards
>>
>> *Jacob Pøhlsgaard*
>> Department of Biochemistry and Molecular Biology
>> SDU, Denmark
>>
>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:56 CST