RE: Unrealistically High Keq in eABF Seperation

From: Ishaan Roy (royi_at_uchicago.edu)
Date: Wed Sep 08 2021 - 13:46:47 CDT

Hello,

Thanks to your help, Chris, I found that the sampling was indeed not adequate, and the PMF values were being inflated. However, the root problem still stands; the PMF (and the PMF gradient) continues to heighten to an unrealistic value, even long after the proteins should not be interacting anymore. I am beginning to suspect that something is hindering the separation. Is there any way to help narrow down what this might be?

Thanks,
Ishaan

From: Chris Chipot <chipot_at_illinois.edu>
Sent: Thursday, August 26, 2021 1:16:14 PM
To: Ishaan Roy <royi_at_uchicago.edu>
Subject: Re: namd-l: Unrealistically High Keq in eABF Seperation

Hi, Ishaan,

hard to say without specific detail. Still, please, check https://urldefense.com/v3/__https://doi.org/10.1080/08927022.2020.1775222__;!!DZ3fjg!r7ZuUmGfAnbBUxIgFbhBlrgDXTBDIMVRIuI9iAy7EthweALME3PgIn6sn1ZKEg-jJw$ , which might give you some pointers.

Best regards,

Chris

On 8/26/21 11:08 AM, Ishaan Roy wrote:

Hi Chris,

I checked over the first and third points you mentioned (they appear fine), but could you elaborate on the second point? Currently fullSamples is set to 5000, would this be considered sufficient in a lipid membrane setting?

Thanks,

Ishaan

From: Chris Chipot<mailto:chipot_at_illinois.edu>
Sent: Thursday, August 26, 2021 2:42 AM
To: namd-l_at_ks.uiuc.edu<mailto:namd-l_at_ks.uiuc.edu>; Ishaan Roy<mailto:royi_at_uchicago.edu>; Peter Freddolino<mailto:petefred_at_umich.edu>
Subject: Re: namd-l: Unrealistically High Keq in eABF Seperation

Ishaan,

there are a number of quantities that you ought to look into. First, you should histogram the difference between the extended variable and the actual variable (from your traj file), and ascertain that it obeys a normal law. Second, you should make sure that your bias is properly applied, that is past a threshold (fullSamples) of accrued force samples guaranteeing minimum non-equilibrium effects. Third, you should make sure that your distanceXY component is correctly defined, with the desired subsets of atoms.

Regards,

Chris

On 8/26/21 12:28 AM, Ishaan Roy wrote:

Hi Peter,

Nothing stands out as weird about the runs, visually or otherwise. The sampling distribution is, I believe, as one would expect, generally increasing for higher values of the coordinate and the trajectory follows a square-root diffusion trend. Thereís no sign of the protein getting stuck anywhere.

Thanks,

Ishaan

From: Peter Freddolino<mailto:petefred_at_umich.edu>
Sent: Wednesday, August 25, 2021 10:34 PM
To: <namd-l_at_ks.uiuc.edu><mailto:namd-l_at_ks.uiuc.edu>; Ishaan Roy<mailto:royi_at_uchicago.edu>
Subject: Re: namd-l: Unrealistically High Keq in eABF Seperation

Dear Ishaan,

If your simulations haven't yet converged, there is really no sense in evaluating the energies completely yet. With that said, how is the sampling looking across the different values of your reaction coordinate? Is the reaction coordinate stuck somewhere or is the system diffusing freely along it?

Have you visually inspected your system to make sure nothing weird is happening?

Best,

Peter

On Tue, Aug 24, 2021 at 7:08 PM Ishaan Roy <royi_at_uchicago.edu<mailto:royi_at_uchicago.edu>> wrote:

Hello,

I am using eABF to separate two membrane proteins to find the free binding energy. The PMF values produced in my runs seem to be incredibly high compared to consistently reaching the 100s order of magnitude. We havenít completed a run that plateaus in free energy yet, but this is an example showing what the PMF looks like.

[cid:part9.F225B075.9737D567_at_illinois.edu]

A maximum PMF this high already produces a value for the binding constant that is on an order beyond 10^100. Besides being high, it might be worth mentioning that the nonbonded energies are very close to 0 near the end of this run, but the PMF gradient doesnít level off at all. Any ideas on what might be causing this? Or is such a value plausible? This only seems to be a problem with the separation runs; my other colvars produce PMF plots that are closer to what Iíve seen.

Thanks,

Ishaan

--
Chris Chipot, Ph.D.
CNRS research director, University of Lorraine
Adjunct professor of physics, University of Illinois, Urbana-Champaign
_______________________________________________________________________
Chris Chipot, Ph.D.
Theoretical and Computational Biophysics Group
3161 Beckman Institute for Advanced Science and Technology
University of Illinois at Urbana-Champaign
405 North Mathews
Urbana, Illinois 61801                            Phone: (217) 244-4361
                             E-mail: chipot_at_illinois.edu<mailto:chipot_at_illinois.edu>
                                     Christophe.Chipot_at_Univ-Lorraine.fr<mailto:Christophe.Chipot_at_Univ-Lorraine.fr>
                             Web:    https://physics.illinois.edu
                                     https://urldefense.com/v3/__http://www.lia-uiuc.cnrs.fr__;!!DZ3fjg!r7ZuUmGfAnbBUxIgFbhBlrgDXTBDIMVRIuI9iAy7EthweALME3PgIn6sn1blMUtJlw$ <https://urldefense.com/v3/__http:/www.lia-uiuc.cnrs.fr__;!!DZ3fjg!oNd5Xch5IACSlLPACt5V7MznhzH0DBZOIV0Ad7knEPsMOq00SM42bqZYLrj44ogkbw$>
                                     http://www.ks.uiuc.edu/~chipot
                                              Fiat justitia, et pereat mundus
                                                  --Heinrich von Kleist
_______________________________________________________________________
--
Chris Chipot, Ph.D.
CNRS research director, University of Lorraine
Adjunct professor of physics, University of Illinois, Urbana-Champaign
_______________________________________________________________________
Chris Chipot, Ph.D.
Theoretical and Computational Biophysics Group
3161 Beckman Institute for Advanced Science and Technology
University of Illinois at Urbana-Champaign
405 North Mathews
Urbana, Illinois 61801                            Phone: (217) 244-4361
                             E-mail: chipot_at_illinois.edu<mailto:chipot_at_illinois.edu>
                                     Christophe.Chipot_at_Univ-Lorraine.fr<mailto:Christophe.Chipot_at_Univ-Lorraine.fr>
                             Web:    https://physics.illinois.edu
                                     https://urldefense.com/v3/__http://www.lia-uiuc.cnrs.fr__;!!DZ3fjg!r7ZuUmGfAnbBUxIgFbhBlrgDXTBDIMVRIuI9iAy7EthweALME3PgIn6sn1blMUtJlw$ 
                                     http://www.ks.uiuc.edu/~chipot
                                              Fiat justitia, et pereat mundus
                                                  --Heinrich von Kleist
_______________________________________________________________________

B93E01B608D94097BD9FC540C8AB8BF0.png

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST