Re: Minimization Question

From: McGuire, Kelly (mcg05004_at_byui.edu)
Date: Tue Nov 13 2018 - 09:39:49 CST

I agree, that would be great to have a sanity check like that. I took me about 3 days to realize I had a QM region embedded in another QM region. This would be great for people just learning how to do QM/MM.

Kelly L. McGuire

PhD Scholar

Biophysics

Department of Physiology and Developmental Biology

Brigham Young University

LSB 3050

Provo, UT 84602

________________________________
From: owner-namd-l_at_ks.uiuc.edu <owner-namd-l_at_ks.uiuc.edu> on behalf of Gerard Rowe <GerardR_at_usca.edu>
Sent: Tuesday, November 13, 2018 6:54:53 AM
To: namd-l_at_ks.uiuc.edu
Subject: Re: namd-l: Minimization Question

I have also made a similar mistake before. Would it be possible for the development team to add some sort of sanity check to the QM region setup phase? In one instance, I had a hard time with wavefunction convergence in a job with two QM regions only to realize later that I had placed an amino acid from QM-1 into QM-2, which was 30A away.

I think it is a safe assumption that nobody would intentionally set up a QM region with a large gap in it, or set up two regions with no gap between them. At least, a warning printed in the output file would be very helpful in troubleshooting QM/MM jobs.

________________________________
From: owner-namd-l_at_ks.uiuc.edu <owner-namd-l_at_ks.uiuc.edu> on behalf of McGuire, Kelly <mcg05004_at_byui.edu>
Sent: Monday, November 12, 2018 9:52:19 PM
To: Brian Radak
Cc: namd-l
Subject: Re: namd-l: Minimization Question

 Thanks Brian, I suspected the a bad choice in the quantum region. Not sure how, but my ligand was put into it's one QM region, and I ended up with two QM regions. The sidechains and waters were one QM region, and the ligand was another QM region within the first QM region. So, it looks like two QM calculations were occurring in the same region. I fixed my config file and my qm pdb to place the ligand back with the other QM region, and everything is working great now.

Kelly L. McGuire

PhD Scholar

Biophysics

Department of Physiology and Developmental Biology

Brigham Young University

LSB 3050

Provo, UT 84602

________________________________
From: Brian Radak <brian.radak_at_gmail.com>
Sent: Monday, November 12, 2018 7:41:26 PM
To: McGuire, Kelly
Cc: namd-l
Subject: Re: namd-l: Minimization Question

That sounds like a bad starting structure or otherwise bad choices in the quantum model and/or quantum region. If minimization is causing your system to distort, you probably have to look somewhere besides your equilibration protocol.

On Mon, Nov 12, 2018 at 8:50 PM McGuire, Kelly <mcg05004_at_byui.edu<mailto:mcg05004_at_byui.edu>> wrote:

  I removed the restraints on the atoms in the QM region. During the last few steps of minimization, one of the carbon atoms on the benzene ring of my ligand ends up near the center of the ring.

  I moved forward, not worrying to much about the minimization result, and ran the annealing step. I can't get this step to finish. Four hydrogen atoms get stretched quite far, and the job crashes because their velocities are too high. Two of the hydrogens are on a nitrogen on my ligand, the third is on an oxygen on my ligand, and the fourth is on an alpha carbon on one of the valine sidechains. What would cause these hydrogens to move so far during annealing?

Kelly L. McGuire

PhD Scholar

Biophysics

Department of Physiology and Developmental Biology

Brigham Young University

LSB 3050

Provo, UT 84602

________________________________
From: Brian Radak <brian.radak_at_gmail.com<mailto:brian.radak_at_gmail.com>>
Sent: Monday, November 12, 2018 12:45:43 PM
To: namd-l; McGuire, Kelly
Subject: Re: namd-l: Minimization Question

If this kind of minimization uses the regular NAMD steepest descent/conjugate gradient minimizer (which I think it does) then you almost certainly should use fewer steps. I would rarely recommend more than 200 steps.

Cartesian restraints (ala the horrendously misnamed "constraints" keyword) are very probably recommended, but maybe not on the QM region?

On Sun, Nov 11, 2018 at 8:47 PM McGuire, Kelly <mcg05004_at_byui.edu<mailto:mcg05004_at_byui.edu>> wrote:

  Usually, during a normal MM simulation, I run the minimization for 1,000 to 2,000 steps. Now, with a QM/MM simulation, using PM7, MOPAC, and GPUs, I can minimize for 1,000 steps in 25 minutes. However, looking at the minimization trajectory, in the QM region some of the sidechain atoms move drastically, causing their bond to stretch really far. I tried restraining the whole sidechain, but the it still happens.

1) Should I not use restraints on the atoms in the QM region?

2) Should I use fewer steps during minimization?

3) Could it be PM7, maybe use a non-semi-empirical theory for minimization?

Same questions for the annealing and equilibration steps...

Kelly L. McGuire

PhD Scholar

Biophysics

Department of Physiology and Developmental Biology

Brigham Young University

LSB 3050

Provo, UT 84602

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2019 - 23:20:21 CST