On Mon, Feb 25, 2013 at 7:31 PM, Marek Maly <marek.maly.ujep.cz> wrote:
> Dear prof. Case,
>
> first of all thank you very much for your comments !
>
> Meanwhile I did some progress. I found out that the solution how to
> prepare/parametrize molecular residuum/molecule which is problematic for
> Antechamber
> due to the unusually high number of bonds (for Antechamber higher than 6)
> might be to use of MOL2+FRCMOD (where Antechamber is not necessary)
> instead of
> PREPIN+FRCMOD (where Antechamber is involved).
>
> If one use MOL2 + FRCMOD there is necessary to solve similar problem as in
> case
> of Antechamber but now just with tleap and as I found the solution is
> probably
> very simple at the end.
>
> It seems (at least in case of Amber12) that only adjusting of MAXBONDS
> symbolic constant (with the actual value 8) in
> AmberTools/src/leap/src/leap/atom.h and recompiling is enough.
>
> prmtop + inpcrd files created successfully, pmemd.cuda seems to have also
> no problems
> with my residuum containing 12 coordination bonds to one atom.
>
> But I have some related additional questions:
>
> #1
> Is this approach really OK (secure) ?
>
I'm not sure. You could always check that the parameters you expect to see
in the topology file are really there. ParmEd and rdparm can help with
this.
You are probably one of a very (very) small number of people attempting
more than 6 bonded partners to a single atom, so you're unlikely to get any
definitive answers. At the very least, you will need to validate your
model.
> #2
> Before I started to search how in leap this bond limitation is implemented
> of course
> that I searched mailing list.
>
> [snip]
> So please can you put some light here and explain
>
> a)
> what is the true/original purpose of "pert true" keyword ?
>
This is very old (in Amber-years) and is now deprecated. This was
originally used by the "Gibbs" program, which stopped being released well
before I joined the project. [1] In the past, TI was performed by
specifying a 'perturbed' residue in the topology file. Basically, the
prmtop stored information about both end states (lambda = 1 and lambda =
0). The parts that differed between the two end states had both sets of
parameters present.
b)
> how is "pert true" related to the problem of leap acceptance
> of more bonds to the given atom than is the actual max. default (defined
> in atom.h).
>
Since both endpoints are present in the topology file, tleap has to support
an atom being bonded to different atoms in each state. Therefore, one atom
_could_ be bonded to 4 atoms in the lambda=0 state, and all 4 of those
atoms could change in the lambda=1 state. Therefore, that one atom needs
to `bond' to 8 atoms (although not more than 4 atoms would be `active' at
any given value of lambda).
What I'm not sure about is whether setting "pert" to `true' simply bypasses
the atom bond limits or whether it does something else, too. Since no
extra sections of the prmtop are printed out, I suspect it just ignores the
hard-coded atom limits, but you should verify this (again, with ParmEd or
something similar).
[snip]
>
> c)
> If "pert true" can really solve the bond problem, is this way more
> appropriate/recommended
> or it is just equivalent to simple redefinition of MAXBONDS constant in
> atom.h ?
>
I'm not sure. I suspect they're equivalent in this particular instance,
but again this is a highly atypical application.
>
> #3
> Regarding substitution of PREPIN file by MOL2 file:
>
> There are some information in PREPIN file which
> are not (at least by default) present in MOL2 like
>
> a)
> IMPROPER record
> which defines groups of atoms (using atom names) for which improper
> torsions have to be used (e.g. aromatic rings).
>
> How to incorporate this important information into MOL2 file ?
> I checked MOL2 format ( http://www.tripos.com/data/support/mol2.pdf )
> and did not found any IMPROPER section.
>
This information should not need to be present --- tleap should
automatically apply known improper torsions that it recognizes.
b)
> In PREPIN file there is also section LOOP where all
> the loops are defined but this info is sufficiently
> substituted in MOL2 file by the complete list of all
> bonds I guess. Am I right ?
>
I don't know what the LOOP section does.
c)
> There are also 3 DU atoms in the start
> of the atom list in PREPIN file defining
> cartesian system. Are these DU atoms
> necessary also in case of MOL2 definition
> of the given residuum /molecule? If yes how to define
> them ?
>
DU atoms are only necessary because the PREPIN uses internal coordinates
(that is the IN part of prepin -- prepc or prepcar files use cartesian
coordinates). The dummy atoms are necessary to uniquely define the
position of the first, second, and third atoms. [2]
Ultimately, there is only one thing that the prepin file has that the mol2
file does not --- connection information. The prepin file allows you to
specify a head and tail atom so that tleap will automatically chain it to
adjacent residues. You can set those atoms by hand, though, and the q4md
folks modified the mol2 file to add this data (see the mol3 documentation
on their website, which is supported in AmberTools 12).
References:
[1]: <hearsay>It had some clever code that was able to efficiently
calculate TIs by only calculating the pairwise differences. However, it
was not well-parallelized, and when pairwise-decomposable methods (cutoffs,
distance-dependent dielectric) began to be replaced by non-pairwise
decomposable methods (Ewald, GB, PB), this approach was no longer
rigorously correct. Therefore, TI was implemented using a dual-topology
approach in sander and Gibbs was dropped. </hearsay>
[2] I suspect this is really just an implementation detail. Since the
positions of DU are arbitrary, so is the position of the first 3 atoms,
technically---at least in Cartesian space. Adding 3 dummy atoms does allow
the first atom's position to be specified by a distance, angle, and
dihedral, just like every other atom.
HTH,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Feb 25 2013 - 17:30:03 PST
