[gmx-users] Re: Free Energy tutorial - choosing number of solvent molecules

[Justin A. Lemkul]

Fabian Casteblanco wrote:
> Hello Justin,
> 
> I am running 2,000,000 time steps for the actual MD run (4,000 ps).
> Depending on how many solvent molecules I start with, I get slightly
> different results.  Do you think its ok to run several different tests
> and take the average?  Or perhaps take the end results of a shorter MD
> run and use those as the starting coordinates for a new run?
> 

Either approach sounds reasonable to me.  Keep in mind magnitude - some of your 
differences were 5 kJ/mol, which may or may not be statistically relevant, 
depending on the value of DG and the resulting error estimates.

-Justin

> Thanks,
> Fabian
> 
> 
> --------------------------------------------------------
> 
> Fabian Casteblanco wrote:
>> Hello all,
>>
>> I'm running the same process from the free energy tutorial by Justin
>> Lemkul...http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin/gmx-tutorials/free_energy/index.html.
>>
>> How did the number of solvent particles get chosen (in the tutorial, 210
>> molecules were chosen)?   I seem to be getting slightly different
> 
> If memory serves, I reproduced what was in the original paper, but do check.
> 
>> results (ranging from as small as 200 J/mol to about 5 kJ/mol depending
>> on how many molecules I choose (ranging for example from 210 ethanol
>> molecules to about 610 ethanol molecules for the largest energy
>> difference change of about 5 kJ).   I keep running tests to see if there
>> is some sort of minimum atom number to get steady consistent numbers but
>> I can't seem to find it.  When I plot for example the bar.xvg &
>> barint.xvg for both sets to see where the lines don't match up, its
>> usually one or two points that differ slightly which cause the free
>> energies in the end to be slightly different.    I seem to be noticing
>> too that the more atoms I use, the free energy gets a little bit lower.
>>
>> Does anybody have any experience with this?
>>
> 
> How long are your simulations?  I have experienced the case (using a
> water-octanol solvent mixture) where the initial configuration made a big
> difference in the result, so longer simulations and multiple configurations for
> the solvent were necessary to get reliable averages.
> 
> -Justin
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================