AMBER Archive (2007)

Subject: RE: AMBER: QM/MM Umbrella sampling

From: Ross Walker (ross_at_rosswalker.co.uk)
Date: Fri Apr 27 2007 - 17:01:46 CDT


Hi Jason,

> &wt type='DUMPFREQ', istep1=100 /
> &wt type='END' /
> DISANG=c_c-11.RST
> DUMPAVE=am1_c_c_vs_t.11
>
> were the restrain file c_c-11.RST reads: &rst iat=3135,3144, r1=1.0,
> r2=2.3, r3=2.3, r4=4.0, rk2=100., rk3=100., /
>
> My question pertains to the restraint file. Are these reasonable
> values for r1 and r4 to obtain a harmonic biassing potential, with a
> minimum of 2.3 Å. Is the force constant defined here as 2 X
> 100 = 200
> kcal mol-1 Å-2?

In terms of the restraint 200 Kcal/mol/a^2 would seem a bit too high to me
but then this is for a bond. The biggest issue with umbrella sampling is you
don't know what the restraint should be until you have done the simulation
:-(... But typically if you set the restraint too large you end up with very
narrow histograms... This means that you have to use many more windows to
get sufficient overlap. However, when in at close distances if you don't use
a large enough restraint then you never sample in close enough - essentially
your histogram has a large tail on it to longer distances and a very steep
gradient at small distances. So typically what you would do is use large
restraints at short VDW separations and then as you move further away so you
can reduce the size of the restraint and also have a bigger distance between
windows. So really I think the best thing you can do is run some short runs
on a few subsets of the windows and see what the overlaps are like. If they
look good then you can extend to longer runs to get good statistics. If they
initially look bad then you can try a different restraint force...

I am not an expert in umbrella sampling though so don't entirely take my
word for it - perhaps others can comment here.

My initial approach here though would be to to try a value of about 50.0 or
so and run the close in windows for say 5ps or so, histogram it and see what
it looks like. Then if the mean position is a long way from where you want
it to be - e.g. say the equilibrium was a 3.0 angstroms and you were trying
to push it to 2.0 but with a too weak restraint such that you ended up with
a histogram centered on 2.8 angstroms then of course increase the restraint
and repeat. If instead you get a histogram centered on where you want but
extremely narrow then decrease the restraint...

So I hope this helps, good luck... I would be very interested to hear how
the QM/MM code performs here.

All the best
Ross

/\
\/
|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

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