[gmx-users] Some questions about free energy calculation
David van der Spoel
spoel at xray.bmc.uu.se
Tue Jan 15 19:29:40 CET 2002
On Tue, 15 Jan 2002 nicolis at guideo.fr wrote:
>Many thanks to both Davids :-) for the answers.
>I am in the reading of a lot of papers.
>Let me ask a little more questions as it is the first time I try the FEP
>I see in the literature that in general we equilibrate the system a few ps for
>each lambda value and than collect the data for the averages.
>In gromacs however, if I define a init_lambda, delta_lambda and nsteps, I
>see that lambda changes by delta_lambda at each step, going from
>init_lambda to init_lambda+delta_lambda*nsteps.
>I understand that if I choose init_lambda=0 and delta_lambda=1/nsteps I
>change from topology A to topology B, but without prior equilibration for
>each lambda value and no "data collection" at each lambda. Is it a
>different approach or it ends to the same DG if the delta_lambda is small
It is better to set init_lambda to a distinct number of values (say 21, 0,
0.05, 0.1 .. 1.0) and delta_lambda to 0.0. You then do each run for say
200 ps and check the curves of dV/dlambda (extract them with g_energy).
You can for each point separately determine average and std. dev. and
check equilibration. If you do 200 ps, you can e.g. use the first 50 ps as
>Is the DG the -kT ln<e^(E/kT)> computed by g_energy with the option -fee
>for the dVpot/dlambda? And I have to apply the Born correction on that
>value, am I correct?
>And the -kT ln<e^(E/kT)> computed by g_energy with the option -fee for the
>dEkin/dlambda should be zero?
You have to integrate the points manually (or e.g. using xmgrace).
Dr. David van der Spoel, Biomedical center, Dept. of Biochemistry
Husargatan 3, Box 576, 75123 Uppsala, Sweden
phone: 46 18 471 4205 fax: 46 18 511 755
spoel at xray.bmc.uu.se spoel at gromacs.org http://zorn.bmc.uu.se/~spoel
More information about the gromacs.org_gmx-users