On Fri, Sep 14, 2012 at 3:49 PM, Fabrício Bracht <bracht.iq.ufrj.br> wrote:
> Hello. I have been trying to figure out why my simulations of a zinc
> protein with a water molecule bound to the zinc atom were failing
> without any warning. I think I found out what the problem is. While
> inspecting the the catalytic site (the residues from the active site
> were parameterized using the MTK++ procedure. To see more details on
> this, see http://archive.ambermd.org/201208/0174.html ), I discovered
> that the angle between the H1-O-H2 atoms of the water molecule was
> practically zero. I mean, they really collapsed onto each other. I
> think the only thing that was holding them from actually collapsing
> was the small vanderwalls parameters I introduced in order to solve
> another problem regarding the charge distribution...etc (not really
> the discussion here). The problem is that the parameters for this
> particular angle is described in the .frcmod file. The line :
>
> HW-OW-HW 103.000 104.520
>
> should have worked, right?
> When I check the parameters of the prmtop file using parmed.py, I find
> that there are no angle parameters for these 3 atoms. None whatsoever.
>
This is expected for many water models (including the TIP series). There
should be 3 geometric parameters holding the water together (since there
are 3 atoms). One option is to set bonds between the O and both H atoms
with an angle connecting them (which seems 'natural' given the bonded
structure of water). Another option, equally valid, is to define bonds
between the O and both Hs and to define a bond between the two Hs,
eliminating the need for an explicit angle term. The 3-bond model is the
path that is chosen for many water models. (This is also why water
molecules appear as a closed triangle when viewed through VMD using the
topology file).
I strongly discourage you from adjusting the water parameters. Note that
these are rigid water models, meaning they were parametrized with (and
never intended to be used without) SHAKE (or SETTLE, which is the analytic
constraint solution for waters). As long as SHAKE is used, what you are
observing should *never* happen, since the H-H distance will be constrained
to its original distance. Did you turn off SHAKE?
I tried introducing them with parmed.py by using the command
> setAngles, but when I try to save the modified prmtop file I get:
>
> Outputting Amber topology file 1g5c_pre1.prmtop
> MoleculeError: Molecule atoms are not contiguous!
>
> Does parmed.py have problems with solvated systems?
>
It's fine with solvated systems as long as that solvated system is not
corrupt :). This is a known LEaP bug that arises when separate chains are
connected via a 'bond' command. The way the prmtop is constructed, all
atoms connected in a bonded network should appear continuously in the
topology file. However, LEaP does not re-order atoms after bond commands
are executed. Which means the ATOMS_PER_MOLECULE and SOLVENT_POINTERS
sections are incorrect. But the error is subtle and does not seem to
destroy your simulation...
Then I tried introducing restraints while doing the tleap routine, but
> since I am using pmemd.cuda, pmemd kindly told me that restraints are
> no longer supported by pmemd:
>
Or sander.
>
> ERROR: PMEMD no longer allows constraints in prmtop!
> | The prmtop must have mbona=nbona, mtheta=ntheta, mphia=nphia.
>
> How can I solve this?
>
Use SHAKE or pick a flexible water model.
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 Fri Sep 14 2012 - 14:00:04 PDT
