VMD-L Mailing List
From: Bennion, Brian (bennion1_at_llnl.gov)
Date: Tue Apr 21 2015 - 16:09:46 CDT
Thanks for the reply. Yes I had previously found a water that had been placed "inside" my molecule. That wasn't going to work.
The DONOR vs ACCEPTOR hint was a good one and helped me catch one issue. I ran both calculations on my molecule and indeed the original guess was incorrect for one of the interactions based on the energetics.
I ended up just clipping off the oxime part of the molecule and parameterizing it separately and that worked out ok.
Again, thanks for jogging the thought process.
From: pawel.kedzierski_at_pwr.edu.pl [pawel.kedzierski_at_pwr.edu.pl]
Sent: Tuesday, April 21, 2015 1:04 AM
To: Bennion, Brian
Subject: Re: vmd-l: fftk water interaction questions
W dniu 20.04.2015 o 21:18, Bennion, Brian pisze:
I have been parameterizing an overall positively charged molecule (scaling=1.1 not 1.16, DistanceWeight=0.8). Most of the compound was described with very minor penalties from the cgenff.paramchem.org server.
When I start the charge optimization for the sections of the molecule that need more validation I am finding that a couple of the atoms are not converging in the MM calculations (objective function is > 1000 at the end of the optimizations).
That is I two atoms for which the final distances in the MM calculations peg to the upper and lower shift boundaries. Increasing or decreasing these two values just moves the MMDistDelta to the new value that was set in the constraints section of the fftk window.
Rather as a general comment, if the QM minimum is shallow (e.g. dominated by dispersion interactions) it may not exist in MM unless you adjust the Lennard-Jones parameters, especially epsilon. But this would apply to atoms with small charges and You have a charged system.
I checked the qm log files that describe these two interactions and they terminated normally with no obvious error. I have looked at the structures from the log files and they appear to settle ok.
I do note that for one of these two atoms, a nitrogen in (-C=N-O-), shows conflicting N...H-O-H distances. Loading in the gaussian log file, the converged distance is 2.46A. However, in the output from the chargeopt.log the QM distance is reported as 4.031A.
You should always check the structures in the GAU files generated by FFTK before calculations. I find it quite common that the initial position of water conflict with some other atoms of the parametrized molecule. In many cases, you may tinker with internal coordinates of the water in the GAU file to make this water point to some other direction. Note that what is optimized in these jobs is only the distance and rotation of this water, and the interaction at this optimized distance should be dominated by the atom where this water is placed and not by some other. I think that even positioning this water at some skewed angle but further from other atoms is better than omitting such data at all.
It is also worth noting that FFTK is rather simplistic about what it thinks is acceptor or donor atom. I would advice to check this, too. Nitrogens for that example are by default considered acceptors but if you have positive Mulliken charge at this nitrogen it most likely will interact better with oxygen ot the water, so it should be put on the "Donors" list in FFTK.
One caveat is that the water in this particular interaction is relatively close to an intramolecular oxygen atom (hbond=2.18A.) I am not sure how much this is affecting the calculation at the N atom.
Do I toss these data points out or is there something more subtle that I can do so solve this problem?