[gmx-users] FEP trajectory errors

[Robert Johnson]

Hello everyone,
I'm computing the free energy of binding of a DNA base on a carbon
nanotube. I think it's a pretty simple calculation and I'm proceeding
in a very standard way. This is what I'm doing:

I have the optimal orientation of the base on the nanotube. I'm
constraining the positions of the base atoms with a soft harmonic
potential.

I then am running two different FEP calculations: one where I turn off
the charges on the base atoms and a second where I then turn off all
the lennard-jones parameters for the base atoms. For each of these I
use the following:
delta_lambda = 0
sc_alpha = 0.7
sc_power = 1
sc_sigma = 0.3

I run a series of trajectories at constant lambda values from 0 to 1.

However, I notice some problems with the trajectories when I turn off
the LJ parameters. As lambda is varied from 0 to 1, it seems that the
position restraints no longer are being applied. Additionally, the DNA
base geometry starts to become severely distorted at lambda values
greater than about 0.6. This happens despite the fact that I am not
perturbing the internal bonded interactions of the base. Here is a
sample of my topology (included just the first line of each section):

[ moleculetype ]
; Name            nrexcl
 dg-disappear             3

[ atoms ]
1       P       1       DG      P       1       1.1659  30.9700
 DUM     0.0000  30.9700
...

[ bonds ]
1       2       1       0.1480  439320.0000     0.1480  439320.0000
...

[ angles ]
1       4       5       1       120.5001        836.8000
120.5001        836.8000
...

[ dihedrals ]
1       4       5       8       1       0.0000          1.6039  3
 0.0000          1.6039  3
...

[ position_restraints ]
1       1       100     100     100     100     100     100
...

So you can see that I am not perturbing the bonded interactions as I
have just copied the same values from topology A.

Also, I am experiencing additional problems when lambda=1. After about
2ns, all the motion in the system begins to freeze and all the atoms
simply vibrate about a fixed position. Soon after that the simulation
crashes.

Can anyone comment on what's going on here?
Thanks,
Bob Johnson