On Thu, 2013-10-31 at 16:20 -0400, Adrian Roitberg wrote:
> Hi All,
>
> I found a peculiar behavior that I need help solving.
>
> 99SBildn has created some new atom types that seems to conflict with
> other names. In particular, ILE has a new C3 type, associated with the
> CG2 carbon atom.
>
> Tleap assign that CG2 as C3, and if one uses GB, it assign a value of
> the Born radii of 2.2, clearly too large.
>
> Same exact setup with ff99SB, assigns a CT type to CG2, and a born radii
> of 1.7
>
>
> The conflict seems to be with C3 atom names being present in some
> version of the UA FFs and also in nucleic acids.
>
> I do not want to start a flame war about how to solve the overall atom
> name issue, but rather to see if someone can come up with a clever
> solution short term (besides editing the prmtop !) and to tell anyone
> that has used ildn with GB to be very careful and check their results...
This was actually brought up over a year ago on the main mailing list
(to which I responded, of course ;)). Here is the relevant link:
http://archive.ambermd.org/201209/0167.html [my response is later at
http://archive.ambermd.org/201209/0216.html].
The current workaround is to assign the correct radii using ParmEd,
which can be done pretty simply. While this is technically 'editing the
prmtop', it is automate-able and not error-prone (outside of forgetting
to do it). The 'best' solution is to assign the radii correctly in tleap
in the first place, and to do it based on chemical knowledge
(connectivity, atomic number, charge, 'hybridization', etc.) rather than
a fragile definition based on (arbitrary) atom type names.
For the time being, the intrinsic radii can be corrected using ParmEd as
follows:
parmed.py -p my_prmtop.parm7 << EOF
setOverwrite
change RADII :ILE.CG2 1.7
outparm my_prmtop.parm7
EOF
Hopefully this is a good enough temporary band-aid until someone
overhauls the radii assignment routines in tleap.
All the best,
Jason
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Nov 01 2013 - 10:00:03 PDT
