[gmx-users] problem with free energy calculations

[Michael Brunsteiner]

Dear all,

A couple of weeks ago i reported a problem I 
encountered with free energy calculations. I've
figured out by now what the problem was (a mistake in
the input, as usual ;) ... however, while searching
for the solution i made a couple of observations
that might be interesting for others (including the
developers). they are listed below.

1) the result of the electrostatic contribution does 
depend on what soft core parameter p i use.
If I use p=1 the results are wrong (10-15 kJ/mol off the
literature value), with p=2 i can reproduce literature 
values (+/- 1 kJ/moil). It might be that with p=1 it simply 
takes much longer to converge, but then in both cases the 
error-bars look quite reasonable, and also from visual 
inspection of dG(t) the results seem to have converged.

2) i get the same results when using md+NH thermostate
with two thermostates, or langevin dynamics (sd), the 
error bars are somewhat larger with sd though.

3) i get the same results when doing slectrostatics
and VdW seperately or in one go, (provided i use p=2).

system/protocols: 462 TIP3P water in a cubic box, 12 A cut-off,
PME with nk=32 or 64. free energy from trapecoidal integration of
<dU/dlam> from 21 windows, at least 100 ps per window,
sc_alpha = 0.5-2.5, sc_p= 1 or 2, sc_sigma=0.3.
error-bars taken to be the RMSD of dGs obtained 
from 10ps/window subsets.

for what its worth ...

mic




 
____________________________________________________________________________________
Never Miss an Email
Stay connected with Yahoo! Mail on your mobile.  Get started!
http://mobile.yahoo.com/services?promote=mail