[gmx-users] Stange results of bonds fep

[Oleg Stroganov]

Dear all,

I'm trying to calculate dG of transition between alpha and pi
conformation in some part of my protein. I indended to use fep and
perform this conversion in three steps: first, I want to fix alpha
conformation gradually by indroducing bonded interactions between
i'th H, i+4 and i+5'th O, then change equlibrium distances of these
bonds from alpha- to pi- state and finally remove bonds.

Surpringly, my system behaved in quite unusual way : at the first
steps of simulation (I used slow growth fep) distances between
perturbed atoms increased dramatically from 0.2nm and 0.45nm
(which are equlibrium distances for alpha-state) to 0.7 nm and
even further. No increase, however, was mentioned when fep was turned
off.

I succeeded to reproduce this strange behaviour on smaller systems by
cutting the piece of the protein, converting it into poly-cly chain
and performing the same simulation with following parameters f A and B
state:

[ atoms ]
...
    16   opls_236      2    GLY      O      4       -0.5    15.9994   ; qtot 1
...
    23   opls_236      3    GLY      O      6       -0.5    15.9994   ; qtot 1
...
    46   opls_241      7    GLY      H     13        0.3      1.008   ; qtot 0.8
...
[ bonds ]
...
   23    46     1     0.20    0     0.20    1000
   16    46     1     0.45    0     0.45    1000

At the and of 50 ps bd simulation from lambda=0 to lambda=1 distance
between 23 and 46 atoms increased from 0.196 to 0.448 nm, and distance
between 16 and 46 atoms increased form 0.496 to 0.73 nm.

Can anyone suggest, why results of fep are so strange? Is this a bug
or I'm missing something?

P.S. I run all these tests with gromacs 3.3.1