VMD-L Mailing List

From: Yu Fang (y.fang2_at_ugrad.unimelb.edu.au)
Date: Mon Dec 14 2009 - 05:40:47 CST

Hi Sarah

Thanks for your information. I think I can clarify things a bit. Up until
this point I have been doing the parametrization of my ligand entirely by
hand. Where required CHARMM force field parameters for the ligand did not
exist in the default parameter files supplied with namd, I took the
closest match by chemical composition.

Where close matches did not exist (for a specific dihedral) I was told by
my supervisor to perform a rotation of this dihedral through 360 degrees,
10 degrees at a time, and performing a qm based minimisation of each of
the 36 resulting conformations (with the dihedral kept fixed). This was
done in Maestro using a qm package called Jaguar.

Doing a plot of the energies of the 36 conformations gave a roughly
sinusoidal graph which represents the torsional potential energy of this
dihedral.

My supervisor then instructed me to adjust my estimated CHARMM force field
parameters until minimization of the 36 conformations using these
parameters produced energies which roughly matched the output of the qm
based minimization. When this is achieved, the CHARMM parameters are a
good representation of the qm model.

However you can imagine the difficulty of trying to adjust the CHARMM
parameters to fit the qm model. (There are actually 4 CHARMM dihedral
parameters involved, plus electrostatics & VDW forces, and the qm graph is
of multiplicity 3 with 2 uneven peaks...) In any case, I was hoping the
paratool would take a qm minimisation/optimisation output of some kind,
and be able to approximate this using CHARMM force field parameters. If
you read the documentation for the paratool, this is what it aims to do.

So contrary to what you mentioned I believe that at least one qm based
calculation is necessary. What I am unsure about is which one(s) out of
the possible 3. The qm geometry optimization will likely change your
molecule (you can see it is different in vmd) because it is using qm model
to optimize/minimize it. I'm not sure what you mean by using "analog
parameters", nor where they come from.

Also on Friday I spoke briefly to someone who had used paratool in the
past and got it to produce CHARMM parameters from qm output but he claimed
the parameters looked suspicious. He modeled a benzene molecule and
running dynamics using the CHARMM parameters made from paratool caused the
benzene ring to bend and flex unexpectedly.

Yu

> dear Yu Fang,
>
> i mean after using QMTool (autogenerating coordinates and writing
Gaussian
> input file). and no, the geometry optimization, the single point calc or
the
> co-ordinate transformation aren't necessary. i didn't do them and i
didn't
> any errors.
>
> specially if you're transferring your ligand to VMD after calculations
and
> processes from another softwares, don't do geometry optimization. it may
change your molecule. (i should admit I'm not professional so I'm not
sure
> about anything i say!!)
>
> i used analog parameters to assign necessary values. for the error of
"expected floating-point number but got " others should help. i myself
received an application error saying it can't recognize the DEFAULTER
command. you'd better use analog parameters and be careful that even if
one
> parameter remain unassigned, it won't appear in parameter file and in
simulation you will receive errors.
>
> if you have N N bonds or angles containing N N or some N N N diheadrals,
you
> can't find any parameter in analog lists. so i myself wrote down the number
> of them the bond, angle or diheadral and added them to the final
parameter
> file manually.
>
> that was all i knew. and i should thank for one of your questions. it
helped
> me fix another error.
> Feel lucky,
> Sarah
>