Re: ABF Free Energy Calculation

From: Joey Farrell (
Date: Wed Oct 05 2022 - 16:16:26 CDT

Hi Haohao,

Thank you very much for the reply. I also appreciate the BFEE2 automation tool you have.

However, because I am running eABF on two helical, POPC membrane embedded proteins, I was running more than one unbound state simulation (i.e., not just unbound RMSD) since I do not have full analytical/theoretical degrees of freedom for tilt/spin angles given placement in the membrane (versus vacuum). Is this logic sound? In such a case, it seemed to make sense to continue layering restraints in standard ABF fashion, so that the first unbound simulation (uRMSD) has no restraints, then the next has a single restraint (just uRMSD restrained with the same k_RMSD=10), and so on.

If you have any suggestions or comments for such a system, it would be very much appreciated.

Thank you for the help.

From: <>
Sent: Tuesday, October 4, 2022 1:43 AM
To: namd-l <>; Joey Farrell <>
Subject: Re: namd-l: ABF Free Energy Calculation

Hi, Joey,

Generally, you may want to read this paper (;!!DZ3fjg!4mAIlRUN1gCmsJ6NDw1yJOhdGwKyHogSQ9LX_ce6sc8pJTNxm3-Y10Ue5assyx2eL5tJWvfrfrzli-ClXyQjoAvR$ <;!!BpyFHLRN4TMTrA!_12-J8_ataj1blvaYQPJYfgzoe6Ji2CEhdtxDDdy0BTKb47CZEgrZ5m95d-uPst-ISSXZXZoH6F9BGjUdKMI$>) and use the automated tool for the task.

Regarding your question,
(1) You should reasonably set the range of the CVs to guarantee sufficient sampling of all bins. Otherwise the simulation may not be converged.
(2) In the unbound-state simulation, there should not be a harmonic restraint. See input files of step 8 generated by BFEE2.
(3) I do not understand the question. k can be an any value. The contribution of harmonic restraints, which is added one by one, will be calculated by post-treatments.



From: Joey Farrell<>
Date: 2022-10-04 14:12
Subject: namd-l: ABF Free Energy Calculation
Hello all,

I have three questions regarding how to properly use ABF to calculate Keq after reading the tutorial and searching the mailing list.

  1. When calculating ∆G via discrete integral, should one only consider bins with a sufficient zcount, or ought all bins on the reaction coordinate contribute? If there is a minimum/threshold zcount, is it a function of the fullSamples parameter, or is it a general, approximate threshold?
  2. In the ABF tutorial, the "unbound" state simulations have colvars which refer to the "bound" state structure as the reference structure. Should the centers of the harmonic restraints be updated such that there is no harmonic potential in frame 0 (i.e., an RMSD restraint originally with center 0 now has non-zero center since the reference structure is not itself)?
  3. Each colvars forceConstant is only realized by a proceeding simulation (i.e., there is no value for "k" until the next colvar's simulation, which then uses the previous as a harmonic restraint). What, then, is the "k" value for the last colvar that is run, since it is by definition never used as a harmonic restraint in another colvar's simulation? It seems it cannot be the case that k=0, as inductively you would never need the last simulation (since ∆G contribution is 0 for k=0 in my understanding).

Thank you very much; I greatly appreciate any help with these questions.


This archive was generated by hypermail 2.1.6 : Tue Dec 13 2022 - 14:32:44 CST