From: Gumbart, JC (
Date: Thu Feb 18 2021 - 18:12:36 CST

As I recall, both the QM and MM energy sets get independently shifted so that their global minimum is set to zero. So it sounds like you have one anomalously low conformation messing up the rest. While it would be good of course to understand why, you might try removing that scan and seeing how it looks then.


On Feb 15, 2021, at 4:50 PM, Smith, Harper E. <<>> wrote:

Hi JC,

Thank you for your reply. My understanding is that the energy cutoff only affects which QM energies are fitted -- is that right? I limited my QM scan to only low-energy regions, so it stays under about 15 kcal/mol globally. Can I do something similar to help my MM energies? They are generally around 100-250 kcal/mol (max ~300). Only 5 MM frames have energy under 10 (and they are consecutive). Despite that, I loaded the trajectory of the MM scan, and it looks normal enough to me. I scan 5 dihedrals in this process, yet the energy only approaches 0 once. Shouldn't it be near 0 at the start of each scan?

This may not be related, but some of the early PSF files generated by Molefacture had duplicate entries for bond definitions, and strange bonds like "1 3 1". I think these were causing errors in FFTK, so I deleted all duplicate entries and "nonsense" bonds.

Happy to send any files that might help reveal my problem.

Harper Smith
From: Gumbart, JC <<>>
Sent: Monday, February 15, 2021 10:55 AM
To: Smith, Harper E. <<>>
Cc: VMD Mailing LIst <<>>
Subject: Re: vmd-l: ffTK torsion scan?

Hi Harper,

I think most people doing parameterization these days use dihedral scans. FFTK also sets an energy cutoff for fitting purposes (about 8-10 kcal/mol by default as I recall). Can’t you just fit the part before clashes increase the energy significantly?


On Feb 11, 2021, at 7:30 PM, Smith, Harper E. <<>> wrote:

Hi all,

I am trying to use ffTK to parameterize a connection between two relatively rigid groups (a ribose and a guanidinium group). I don't think I can remove anything else from the system -- it's already as basic as possible. In the dihedral scanning portion, there are big clashes and they cause large energies in the MM portion. If I keep kmax at 3.0, the smallest RMSE I can get is around 60.

The ffTK paper mentions that estimates for dihedral force constants can come from vibrational analysis. Is this the best option for a system like mine? If so, could someone describe what this calculation would entail? ffTK has been an introduction to QM calculations for me, so any details or suggestions will be helpful!

Harper Smith