Methane in water box FEP

From: Sophia Kondratova <x357l_at_unb.ca>
Date: Fri 23 Nov 2001 15:33:25 -0400

Hello,

I am trying to do the "methane in water box" example, posted on the website
and I am having trouble doing the free energy calculation part. I followed the
instructions on the website and used the same parameters for minimization and
molecular dynamics:

1)a minimization using constant volume periodic boundary conditions (using
sander/pme; pme was not used in the web example version)

2)an md run to equilibrate, using constant pressure periodic boundary
conditions (again using sander/pme) 5000 steps with initial temperature 0.0
and the target temp 300.0 K

3)another md run again with constant pressure pbc (same as above, but taup =

4)free energy perturbation, time step of 0.0005 psec, 10000 steps per window,
21 windows, lambda 0.05 initial temp 0 (i did do it with init temp 300 too)
as well no pme was used with gibbs

Once these are done, looking at the total free energy, I get a big difference
between the forward and reverse free energy, mainly

/FREE ENERGIES/
FORWARD # STEPS EQUILIBRATION DATA COLLECTIO
WIND LAMBDA dLAMBDA NR EQ COLL FORWARD REVERSE FORWARD REVERSE

16 0.250000 0.050000 0 2000 8000 0.9259 -1.0378 1.2268 -1.8746
17 0.200000 0.050000 0 2000 8000 2.2268 -1.4965 1.4176 -2.0974
18 0.150000 0.050000 0 2000 8000 0.2754 -81.3171 0.1307 -81.5529
19 0.100000 0.050000 0 2000 8000 0.2030 -0.2077 -0.1744 -164.0198
20 0.050000 0.050000 0 2000 8000 0.1261 -8.3158 -35.9359 -105.6102
21 0.000000 0.050000 0 2000 8000 1.7114 -0.7587

*SUMMARY*
FREE ENERGIES
# STEPS DATA COLLECTION
AVE TOT AVE AVE TOTAL TOTAL
WIND STEPS LAMBDA dLAMBDA NR EQ COLL FORWARD REVERSE
21 210000 1.00-0.00 0.047619 0 2000 8000 -30.3683 -359.0144

For some reason, near the end of the simulation the forward and reverse
free energies are completely different, unlike the rest of the simulation,
where
the forward and reverse free energies are comparible. I decreased the time
step,
and I did do the FEP for 20000 steps md per window (longer simulation time),
but that did not help iether.

The only difference between my input files and the ones on the webpage is the
fact that I am using sander/pme for min and md.
Any suggestions would be appreciated. Also if anyone knows of other good
starting examples to try for free energy calculations on a solvated system
using amber, this would be very helpfull.

Sophia Kondratova
University of New Brunswick