AMBER Archive (2005)

Subject: Re: AMBER: new tutorial available

From: Julien Michel (j.michel_at_soton.ac.uk)
Date: Wed Jan 26 2005 - 07:48:55 CST


Hi,

> Dear Mr. Case,
>
> Thanks for the new tutorial. I have a couple of questions to ask.
>
> 1. The transformation is from toluene -> nothing, and in order to do that,
> u do a 2-step MD simulation; one called as 'charging' and the other as
> 'disappearing'. For charging, the free energy difference is calculated as
> 2.35 kcal/mol, and for disappearing, the free energy difference is
> calculated as (for klambda=6) 2.7 kcal/mol. So, the total free energy
> difference of toluene -> nothing transformation is 2.35 + 2.7 = 5.05
> kcal/mol, am I right?
>
> 2. For the charging case, u are doing a 3-point Gaussian Integration,
> while for disappearing case, u are doing a 14-point Gaussian Integration.
> Is there any optimum number of lambda values for different structures at
> interest? For instance if I do a simple perturbation on the amino group of
> an RNA molecule, how can I decide how many lambda values I should use?
>
    Unfortunately, it is difficult to know a priori how smooth the PMF is
going to be. You can often get a rough guess by running a short simulation at
each lambda and see what the PMF looks like.

> 3. For the disappearing case with klambda=6, you have chosen lambda values
> of 0.90, 0.92, 0.94, 0.96 and 0.98. Any reason for that?
>
   My guess is : If you check the values of the gradients at these
windows, you will probably see that they are large, and therefore you need to
do get more gradients in this region to get a good profile for the PMF.
I do not know which mixing scheme was used for TI in this tutorial, but there
are often difficulties arising when you get close to the end state because the
LJ term is very 'hard' and you can have a pretty big change in LJ when going
from lambda 0.99 to lambda 1.00 which is going to affect your sampling.

> 4. In the frcmod file which describes the parameters for the dummy atoms,
> you did not use zero values for harmonic force constants of the bonds and
> angles, and the barrier heights (PK). If the atoms are disappearing, why
> not choosing them as zero?
>
        To understand why the hard degrees of freedom cannot be coupled to lambda,
I suggest you read the following papers :
        J.Phys.Chem A,1999,103,103-118
        J.Phys.Chem A,1999,103,119-136

-- 
Julien Michel
University of Southampton
Department of Chemistry
Dr J. W. Essex group 
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu