Re: [AMBER] interpreting paramfit results

From: Cenk Andac via AMBER <>
Date: Sun, 18 Sep 2022 11:04:19 +0300

Hi Robin,
Thank you for taking your time helping me out.
Yes, I did determine the K value as suggested in the tutorial,
and I used the K value as an input for fitting.
Is that what you meant about " the relative difference in energy between
structures" ?

However, I did not use your bash script to randomly generate coordinates
about the dihedral of interest in order to determine quantum energies. And
now, I have started thinking if I really have to.
What I rather did was, I generated coordinates for incremental rotations
about a torion of interest (5 degree increments between 5 to 360 degree) to
yield 72 frames using the leap module of AMBER v20. I then transfered the
coordinates into gamess input files (yielding 72 gamess inpit file) and
determined single-point energies for each frames. Then I extracted
potential energies from gamess output files and collated them into a
quantum.dat file which I used for fitting.

If additional N is desired (say N=2) , then I would have to determine
another K value for a different dihedral angle about the same atom2-atom3,

On Fri, Sep 16, 2022 at 12:15 AM Robin Betz <> wrote:

> Hi Cenk,
> When you ran paramfit, did you fit "K" first?
> The absolute value of both QM and MM energy doesn't really have meaning
> and the values can't be directly compared.
> Instead, the relative difference in energy between structures should match
> (structure 1 and 2 should be X kcal/mol different in energy, for both QM
> and MM).
> "K" is a fittable parameter that adjusts the origin such that QM and MM
> energies can be directly compared, and it should be fit before
> fitting other parameters, and not fit while other parameters are.
> If you are fitting K as the tutorial describes, there may be insufficient
> degrees of freedom in the available forcefield terms to get a good result.
> You can try adding more terms with different values of N to that dihedral
> and see if the result improves.
> Also I am not sure why paramfit was officially deprecated but as the
> original author, I'm happy to continue answering questions about it on this
> list.
> Best,
> Robin
> On Thu, Sep 15, 2022 at 1:46 PM Cenk Andac via AMBER <>
> wrote:
>> Dear Amber users,
>> I applied paramfit to fit single-point rotational energies about a
>> dihedral
>> angle (of an nt derivative) determined by MM (blue plot in the figure),in
>> which Vn values for the dihedral were zeroed in the force field, to single
>> point QM energies (at the MP2 level) for the rotations about the same
>> dihedral angle (black in the figure). The red curve in the figure
>> represents the fit energies.
>> I used Shaw's RNA force field for parameterization.
>> I see that the energies obtained by MM is about 4 times as much as
>> energies
>> obtained by QM.
>> [image: image.png]
>> My question here is WHY?
>> Shouldn't they be about the same?
>> Please see the plot below after dividing the MM energies by 4. Any ideas?
>> [image: image.png]
>> best
>> Cenk
>> _______________________________________________
>> AMBER mailing list
AMBER mailing list
Received on Sun Sep 18 2022 - 01:30:02 PDT
Custom Search