AMBER Archive (2006)

Subject: Re: AMBER: restraints and constant pressure

From: Thomas Cheatham (tec3_at_utah.edu)
Date: Fri Sep 15 2006 - 01:22:02 CDT

> "Using constant pressure with restraints can also cause problems so
> initially we will run 20 ps of MD at constant volume."
>
> I am working with a system where it is really necessary to use
> restraints on about one third of the solute. I had planned on using
> ntr=1 with a force constant of 10 kcal/mol. Could someone please specify
> what the risks are of combining restraints with constant pressure, and
> how I can minimize those risks?

At issue is that pressure scaling works by shifting (scaling) the relative
positions of each molecule (in older versions when NPSCAL=1 or atoms when
NPSCAL=0) in the box to increase or decrease the size.

To simplify the restraint code, rather than inducing a force (coordinate
deviation) every time the simulation coordinates are scaled (relative to
the restraint coordinates), it was decided to scale the restraint
coordinates as well. This effectively changes the restraint coordinates
at each step (every time the coordinates are scaled). Two problems may
result:

(1) when you restart, and reference previous restraint coordinates (that
were not shifted), you will get an initially large restraint energy due to
the "shift", and

(2) if you are restraining multiple molecules, the relative positions of
the molecules will change as the simulation proceeds. Imagine the default
case where the density of the water is too low initially, the box has to
contract. If you are restraining two molecules, for example the two
strands of DNA, as the run continues, the restraint coordinates will shift
the molecules closer together which will force your MD coordinates to
cause the two strands to get closer. This is undesirable.

In the older DNA tutorial, a discussion was present that provided this
info; to get around (2), I normally edit the prmtop file to coalesce the
first two molecules (i.e. two strands of DNA for example) into a single
molecule (and change the pointers to reflect this).

To get around (1), I normally use the restrt file as the refc coordinates
for a continuation...

So, short summary: if you are restraining a single molecule (like a
single protein chain), you likely will not suffer serious artifact and can
get away with shifting (as per (1) above) noting that upon restart the
restraint energy will be a little higher than expected due to the
shifting. As long as the force constant for NTR is not too high (1.0
kcal/mol is often strong enough a restraint), this shift should not lead
to integration failure.

I hope this helps,

--tom

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu