[gmx-users] dg/dl at lambda = 1

[bharat v. adkar]

Dear gromacs users,

I am trying to use FEP method for the calculation of relative binding free 
energy of peptide-protein complex. I am using gmx-3.3.1.

As per previous discussions on the mailing list, i learnt that i need to 
do two different sets of simulations for my case. first set will be for 
modification of electrostatics and second one for vdw. The second set 
should be started with a structure from the last simulation of first set. 
And every simulation in each set should be run independent of each other 
with em and water-equilibration at each lambda value.

following are the values for peptide only (not in complex) simulation in 
SPC216 water-box...

lambda  dg/dl (avg)     error
0.0     55.777174       0.586923
0.015   57.524631       0.808737
0.03    54.648669       1.03151
0.05    54.833890       1.08532
0.1     52.990600       1.70822
0.2     50.377182       1.74758
0.35    46.725657       0.683906
0.5     37.016530       0.826907
0.65    33.239827       2.03776
0.75    27.867797       0.300761
0.85    15.989920       1.66024
0.9     1.591218        0.924152
0.95   -32.877432       1.58611
0.99   -259.656906      9.28882

error is calculated by using g_analyze with -ee option

i am getting very huge values at lambda = 1.0. as can be seen from the 
values, the steepness of the graph has increased a lot towards end. for 
l=1, dg/dl values go till ~ -2 x 10^11.

what can be wrong? I have used the same version of gmx for the 
tutorial provided by Dill group. In that case, i am not getting any 
problem....

I am completely clueless here... please help..

thanks
bharat