[gmx-developers] Re: Amber to Gromacs forcefield conversion

[Erik Lindahl]

Graham Smith wrote:

>Anton Feenstra wrote :
>
>>I've just now commited my amber2gmx forcefield comversion scripts to the
>>CVS in src/contrib/scripts. 
>>
>
>>Known bugs and problems at this moment are:  - pairs, grompp says: no
>>defaults. That is reasonable, since no
>>pairtypes are defined yet. I have to look up how they are defined
>> in the Amber ff. I think it is simply a scaling with respect to
>> the normal LJ parameters.
>>- impropers: a trial EM run on a peptide results in distorted
>> geomoetry. Nothing definite yet, but it seems the impropers are off.
>>
>
>I did
>pdb2gmx -f 4pti.pdb -o 4pti1.pdb -p 4pti1.top -i 4pti1.itp -ss -ter 
>grompp -f mdmin.mdp -c 4pti1.pdb -p 4pti1.top -o 4pti1.tpr \
>-maxwarn 10000
>which works after a fashion, then I gmxdumped the tpr.
>
>CYS S-S briges don't work yet. 
>I had to rename LYS, LYN -> LYSH, LYS in .hdb 
>

Hi Graham,

I'm looking into similar things for OPLS-aa, so maybe I'll be fixing 
some of this in pdb2gmx.

>
>
>As you say, the impropers aren't right yet - the angle and force constant
>are the wrong way round. More worryingly, these are just the impropers for
>pi-bonding: the impropers that maintain the geometry of chiral centres are
>missing, and I can't see them in the amber force field files either - are
>they generated automatically by amber? Maybe we should just generate one
>for every sp3-hybridized atom (...that doesn't have two or three hydrogens
>on it)? 
>
I don't think you need them with the all-atom force field. At least 
that's the case with OPLS-aa;
since we explicitly include all four atoms around the chiral centers the 
geometry won't distort.
(The energy barrier for L-R conversions is just as high as it is in nature!)

OPLS-aa might be slightly different in that it includes all proper 
dihedrals, though.

>
>My -n and -c.tdb files nearly work, once I renamed peptide hydrogen from
>HN to H. There was a weird error with an N-terminal R, which I suspect
>may be a bug; almost every atom has to be in the [replace] list, and pdb2gmx 
>complained about just the 2nd H atom on every heavy atom:
>WARNING: atom HB2 is missing in residue ARG 1 in the pdb file
>         You might need to add atom HB2 to the hydrogen database of
>         residue ARG
>         in the file ff???.hdb (see the manual)
>WARNING: atom HG2 is missing in residue ARG 1 in the pdb file
>WARNING: atom HD2 is missing in residue ARG 1 in the pdb file
>WARNING: atom HH12 is missing in residue ARG 1 in the pdb file
>WARNING: atom HH22 is missing in residue ARG 1 in the pdb file
>It tolerates the HA on CA though. 
>(they *are* there in hdb and have right name) 
>
I experienced the same problem with an ACE capping residue; I'll check ARG.

Cheers,

Erik