[gmx-users] Free energy calculations (FEP) and soft core potential 1-1-48

[Hannes Loeffler]

On Wed, 15 Jul 2015 16:20:25 +0200
Julian Zachmann <FrankJulian.Zachmann at uab.cat> wrote:

> The simulation time for 1ns is quite short of course but I have
> 'nstdhdl' put to 10 to get a lot of output.

To add to what has been said already:  increasing the output within a
fixed set of time won't give you any more information.  Your data is
probably highly correlated.  You may want to look into a tool like
https://github.com/mobleylab/alchemical-analysis if you haven't done
already and have a look how much data survives the correlation analysis
implemented in there.

Also, check for convergence to see if you have still strongly varying
gradients/energies.  Something as simple as a block analysis can be
very effective.


Cheers,
Hannes.