[gmx-users] Re: alchemical free energy changes upon change in sc_alpha and sc_power

[Ilja Khavrutskii]

I should probably withdraw my question. I did some tests on a smaller
system, i.e., ligand in a cubic water box and the free energy differences
between (1.5, 2) and (0.5, 1) I got are within 1KJ/mol, which might be OK
for the single precision gromacs anyway. In the case of the protein complex,
it appears that I might have driven the system out of its proper
equilibrium.

On Mon, Jul 26, 2010 at 12:00 PM, Ilja Khavrutskii <ivkhavru at gmail.com>wrote:

> Hi,
>
> I have seen previous posts on the sc_alpha/sc_power, but they did not
> answer my question: why does the free energy of an alchemical transformation
> (the work that it takes to change state A into state B in a single leg of a
> thermodynamic cycle) change when I change (sc_alpha, sc_power) from (1.5, 2)
> to (0.5, 1). The A and B Hamiltonians should be identical at lambda 0.0 and
> 1.0 regardless of the values of (sc_alpha, sc_power), I think. I see
> differences of 40 KJ/mol in the work values averaged over 8 independent
> simulations that are converged to at least within 5 KJ/mol for each set of
> (sc_alpha, sc_power) (2 ns MD in each TI window, NPT). I am using GROMACS
> 4.0.5 single precision run in parallel. My system is periodic with a
> protein, ligand an cubic water box and I have dummy atoms only in the state
> B. I would appreciate any hints into this. What am I missing that would
> explain the difference of 40 KJ/mol for just changing parameters from (1.5,
> 2) to (0.5, 1). I am specifically asking about the single leg work (deltaG),
> not the whole thermodynamic cycle (delta-deltaG).
>
> Thanks,
> -Ilja
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20100727/257dc8db/attachment.html>