Dear Junmei,
Thanks for your response. It was very helpful. I still have a few
questions to try to explain the strange results that I'm getting.
First, I realize now that I wasn't using the latest version of Antechamber.
I was using Antechamber 1.27 instead of the version included in AmberTools
1.4. I've now parameterized my molecules using Antechamber in AmberTools
1.4, and now I get the same atom types as what you sent me. This seems to
have fixed the problem I observed where aromatic atoms would be typed as nh,
c2, etc. You've also explained to me that the 'na' atom types observed in
imidazole, pyrrole, etc are actually correct, so I'm not worried about
seeing this now. However, I still am getting molecules with conformations
that seem to be incorrect. Perhaps this is a parameter problem rather than
an atom typing problem?
There are at least two parameters that I think are causing problems.
The first is the torsion:
X -ca-na-X 4 1.200 180.000 2.000
If ca is a pure aromatic carbon, and na is a planar aromatic nitrogen, why
is the barrier to rotation so weak here (1.2 kcal/mol)? This comes up when
I parameterize imidazopyridinium. In my first email, I sent a mol2 file for
imidazopyridine. When I parameterize this, the atoms bridging the ring are
typed as cd and na, and have a rotation barrier of 6.8 kcal/mol. However,
in imidazopyridinium (attached), the atoms are typed as ca and na, so the
rotation barrier is only 1.2 kcal/mol. This allows the two fused rings to
fold against each other rather than staying planar. I've attached an
example conformation (pdb) that I've gotten from doing a vacuum simulation
of this molecule using the GAFF parameters. Is there a parameter problem
here, an atom typing problem, or neither? Why should imidazopyridinium have
such a different barrier than imidazopyridine?
The second problem is with angles. Going back to imidazopyridine, when I
parameterize this using AmberTools 1.4, all of the carbons around the
ring-bridging nitrogen are typed cc, and the bridging nitrogen is typed na.
This is also true in the file imidazopyridine.prepi that you emailed me,
although your molecule seems to have added some phenyl groups. The cc-na-cc
angle in gaff.dat (as well as in the imidazopyridine.frcmod that you sent
me) is specified as:
cc-na-cc 68.940 109.900
This is basically a tetrahedral angle, even though all three of these atom
types are supposed to be planar and aromatic. So, the nitrogen which fuses
the rings attempts to be tetrahedral, which results in the rings folding up
against one another (partially) rather than the entire molecule staying
flat.
One molecule that particularly suffers from this angle problem is
imidazothiazolium. This also contains a cc-na-cc angle despite being fully
aromatic (I think?) and results in bent conformations. I've attached a mol2
file and pdb file from a vacuum simulation of this molecule using GAFF
parameters and the bending of the aromatic rings is very obvious. Is there
something wrong with these aromatic angle parameters?
Thanks for your earlier reply and I hope we can identify what is happening
here and fix a problem if there is one.
Finally, I am going to be on vacation from 9/28 until 10/15. I still want
to get this issue resolved, but I won't be able to reply to emails in that
time period. If you identify a problem or can explain to me to what's
happening, I'll look at it as soon as I get back.
Thanks,
Gabriel
Gabriel Rocklin
UCSF Shoichet/Dill Labs
On Wed, Sep 22, 2010 at 8:56 AM, Junmei Wang <Junmei.Wang.utsouthwestern.edu
> wrote:
> Dear Gabriel,
>
> Thank you very much for your question. The following is my explanation.
>
> In GAFF, we have several types for aromatic/aliphatic properties, from AR1
> (pure aromatic, such as benzene and pyridine), to AR2 (planar systems other
> than AR1), to ... AR5 (pure aliphatic systems). Most heterocyclic systems,
> such as pyrrole, imidazole and the five memberred rings of indole and
> 4-azaindole (excluding the bridge/fused atoms), belong to AR2.
>
> When the protonation state changed, the atom type must also be changed. For
> example, the pyridine N is 'nb' if deprotonated and 'na' if protonated.
> Otherwise, there is no way to use generic torsional angle terms (the
> multiplicities are different for 'nb' and 'na'). By the way, 'na' is the
> planar nitrogen with three bonded atoms.
>
> Did you use the new atom type definition file? In the new version, no 'c2'
> atom type existing any more for AR2 atoms. The 'c2' atom type is replaced
> with either 'cc' or 'cd' and a much higher rotational barriers are used in
> GAFF for torsional angles involved 'cc' and 'cd'.
>
> Do you have probamatic molecules that produce wrong/weird conformation in
> your free energy calculations? If you have some, could you please send them
> to me? I would like to figure out what is the problem.
>
> I have generated the mol2, prepi and frcmod files for the molecules you
> mentioned in your last email. Please let me know if you get a different
> result.
>
> Best
>
> Junmei
>
>
>
>
> ------------------------------
> *From:* grocklin.gmail.com [grocklin.gmail.com] on behalf of Gabriel
> Rocklin [gabriel.rocklin.ucsf.edu]
> *Sent:* Monday, September 20, 2010 3:20 PM
> *To:* Junmei Wang; case; amber.ambermd.org; David Mobley
> *Subject:* Aromatic C and N atom type problems in Antechamber
>
> Dear Junmei and others,
>
> I am trying to parameterize various heterocyclic compounds using
> Antechamber, and I'm getting strange results. Nitrogen and Carbon atoms
> that I think should be classified as aromatic are being classified as
> non-aromatic sp2. This creates a problem for free energy calculations
> because the non-aromatic atoms have very weak dihedral parameters associated
> with them which allows incorrect conformations.
>
> I can't trace all the problems I'm seeing to a single issue, so instead
> I'll list several. In all the examples, I'm using the version of
> ATOMTYPE_GFF.DAT that you send me in February, linked below (is there
> something newer?)
> http://archive.ambermd.org/201002/0314.html
>
> First I have some questions about pyrrole and imidazole.
>
> 1) When I parameterize pyrrole.mol2 doing:
>
> antechamber -f pyrrole.mol2 -fi mol2 -o pyrrole.prepi -fo prepi
>
> The nitrogen is classified as type 'na', which is described in GAFF.dat
> as: "Sp2 N with three connected atoms" which is a non-aromatic atom type.
> Shouldn't this be classified as aromatic? (Pyrrole isn't actually a
> molecule I'm interested in, but this is one of the most basic cases of the
> issue I can find).
>
> 2) When I parameterize neutral imidazole.mol2 in the same way, the Ns
> have two different atom types. The deprotonated nitrogen is recognized as
> 'nc', which GAFF.dat says is aromatic ("Sp2 N in non-pure aromatic systems")
> whereas the protonated nitrogen becomes 'na', again not aromatic. If there
> were a substituent coming off this nitrogen (or both nitrogens, in an
> imidazolium derivative) these would be parameterized such that they could
> bend significantly out of plane of the ring because the N was not classified
> as aromatic. Should they be parameterized this way or is the atom type
> assignment wrong?
>
> Now some issues with fused ring systems...
>
> 3) Indole. The N in indole is classified as "na". Shouldn't this be
> aromatic?
>
> 4) Imidazopyridine vs Indole. This is the kind of fused ring system that
> I'm interested in and where I see real problems. In Imidazopyridine, N1
> (the same N as in Indole, except now deprotonated) is recognized as aromatic
> and classified as 'nc'. But the N bridging the ring systems is the
> non-aromatic "na". Several carbons also change atom type. In Indole, all
> of the carbons are 'ca', 'cc', or 'cd' (all aromatic). In Imidazopyridine,
> some carbons switch to 'ce', 'cf', and 'c2', none of which are aromatic.
> The "na" nitrogen bridging the rings means that the two rings actually have
> the flexibility to bend toward each other, rather than the entire molecule
> staying planar.
>
> 5) 4-azaindole vs 4-azaindolium. In 4-azaindole, everything except the
> protonated nitrogen is aromatic. As far as I know, it should be aromatic,
> but this seems like the same problem as above. However, in 4-azaindolium
> (both nitrogens protonated), several carbons become nonaromatic (c2). One
> of the nitrogens even becomes classified as NH ("Amine N connected one or
> more aromatic rings") which must be because it itself is not recognized as
> aromatic.
>
> The above nitrogen issues seem to all come from any nitrogen being bonded
> to three atoms not being recognized as aromatic. However, this doesn't seem
> to be the full extent of the problem, because even the carbons in
> Imidazopyridine and 4-azaindolium become non-aromatic. My main concern in
> all of this is the torsional parameters because I don't want my aromatic
> fused rings to be folding in on each other if they aren't supposed to be.
>
> I've attached mol2 files of all of these molecules to see if you can
> reproduce what I'm seeing, and in the hope that a fix can be found.
>
> Thanks,
> Gabriel Rocklin
> UCSF Shoichet/Dill Labs
>
>
> ------------------------------
>
> UT Southwestern Medical Center
> The future of medicine, today.
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Sep 27 2010 - 22:00:04 PDT
