[gmx-users] comparing gmx GB energy with Amber11

[Per Larsson]

Hi

A few things:

1. If I recall correctly, the GB-energy in gromacs is split into two parts, GB-polarization and non-polar solvation. Can you check whether this is the case and if the value you report is the sum of those two terms.

2. Try setting all cut-offs to 0 (infinite cutoffs). That will trigger a slightly different part of the code to do the GB-calculation (you should see in your log file references to all-vs-all kernels)

3. Try running in single precision, keeping in mind 1 and 2 above to see if that helps. Also, try setting the GMX_NOOPTIMIZEDKERNELS variable to 1 (at least it used to be called that, someone else can perhaps correct me if that has changed). This will make the code run without the sse-loops.

4. It could be that you are using different radii for the atoms in your system. For amber, are you using the Bondi-radii or something else?

Cheers
/Per


8 nov 2012 kl. 05:43 skrev Sandeep Somani:

> Hi,
> 
> I am comparing single point amber ff energies from gmx4.5.5 (double
> precision) and Amber11.
> 
> All bonded and non-bonded energy terms are in very good agreement (within
> 0.1 kJ/mol) except GB:
> gmx 'GB polarization' = -200 kJ/mol
> amber 11 'EGB'         = -283 kJ/mol.
> 
> I am basically trying to replicate Amber's "igb=5 + large cutoff" in gmx.
> According to Amber manual igb=5 corresponds to "model 2 of [2004 OBC
> paper]".
> 
> I translated this to the following mdp options:
>      ----
> implicit_solvent    = GBSA
> gb_algorithm        = OBC
> All cut-offs (rcoulomb, rvdw, rlist, rgbradii) larger than the size of the
> molecule.
> {gb_epsilon_solvent, gb_saltconc, gb_obc_alpha, gb_obc_beta, gb_obc_gamma,
> gb_dielectric_offset} = corresponding Amber input.
>      ----
> 
> Is that correct?
> 
> I realize that this is as much a question for Amber mailing list, but
> trying here first! Any idea?
> 
> The molecule is ala-12. I'll be happy to send input files.
> 
> Thanks in advance
> 
> Sandeep
> 
> --
> Postdoc
> Wales Group
> Cambridge
> -- 
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists