[gmx-users] free energy calculation: combination of position restraint +refcoord_scaling cause crash during NPT

[Jin Zhang]

Dear all,

We're doing free energy calculation and found some of my lambda job crashes
due to the combination of position restraint +refcoord_scaling COM. All
crashes happened at NPT steps.

By checking each energy term, we found the Position Rest. term is
surprisingly as high as 10^6 kJ/mol, the others look fine. Either turn of
position restrain or refcoord_scaling could avoid the simulation to be
crashed. We tried to see if there's mismatch of either COM or coordinate
and found no direct answer.
Any help would be appreciated!

In the tpr file
      refcoord-scaling     = COM
      posres-com (3):
      posres-com[0]= 3.15034e-01
      posres-com[1]= 4.48110e-01
      posres-com[2]= 5.17730e-01

COM of protein-ligand calculated by t_traj -com -ox
3.02422    4.30169    4.84964

coordinate of 1st atom in tpr: posres_xA[    0]={ 1.39783e-01,
-9.56922e-02,  1.22536e+00}
coordinate of 1st atom in gro:     1MOL     C1    1   3.164   4.206
6.075

The same thing was also found in a normal non-free energy simulations with
combination of position restrain and refcoord-scaling.
Later on, we found some other people also have the same problem when use
that combination to do free energy calculation.
http://comments.gmane.org/gmane.science.biology.gromacs.user/66177
It makes more sense to me to turn on refcoord_scaling while using position
restraint.
Again, any help to understand this would be appreciated!

Regards,
Jin