VMD-L Mailing List

From: Josh Vermaas (vermaas2_at_illinois.edu)
Date: Wed Oct 02 2013 - 09:00:54 CDT

Hi Daniel,

In my experience, the noisiness is caused by small geometry changes
leading to large changes in the electrostatic interactions, which in the
cases I've looked at, does get smoothed out so long as the changes
aren't too large. Your example seems to be a particularly pathological
case (MM-minima don't align with QM-minima), and since what FFTK does
internally is to set the absolute minimum found in the QM- and MM-scans
to 0 energy, no amount of adding reasonable dihedrals will recover
having minima in the right place. One thing you could do is to load in
your dihedral scans (Scan Torsions->Load Dihedral Scan LOG Files) and
see which interaction is artificially too favorable (NAMDEnergy is your
friend here), and work on correcting that. My hunch is that there is too
strong of an electrostatic attraction between two distant points in your
molecule, which just needs a bit of charge reoptimization.

-Josh Vermaas

On 10/02/2013 03:24 AM, Daniel Klose wrote:
> Dear VMD & ffTK Users,
>
> two short questions that puzzle me:
>
> First one regarding ffTK dihedral fitting:
> I used ffTK on several small organic molecules so far and it worked really well!! Only now in one example for some reason I could not find so far, during the dihedral fitting procedure the initial MM energy (MMEi) seems awfully noisy and not a smooth as usually.
>
> See red line in this plot: www.israfil.de/files/ffTK_dihedral_opt_problem.png
>
> So in the end, the final energy profile will also be noisy because the dihedral terms only add smooth funtions.
> What noise/variation level would be normal? What may be the reason for this "noise"?
> I tried repeating the fitting of the bonded interactions, but always got the same results with different initial parameters.
>
> All qm calculation were run with the default level of theory from ffTK and converged fine (normal termination and non-imaginary frequencies in the hessian calculation).
>
>
> The second question is probably simpler:
> I'd like to also compute cmap term, i.e. the correction matrices that give the dihedral-dependent offset energy to be added to the MM energy to better match the QM energy landscape.
> My question is: Should I compare energies for exactly the same stuctures (and equal dihedrals) or should I compare energies between relaxed QM and relaxed MM structures with the dihedrals fixed to the cmap grid values?
> And how should the scaling of the energies be done to calculate the matrices once the target qm data is obtained?
>
>
> Thank you for your help & advice!! :)
> Daniel
>
>