reconstructing PMF from CVs other than progress coordinate

[Paween Mahinthichaichan]

Hello,

I am using ligand's RMSD as the progress coordinate in WESTPA simulations. However, I would like to reconstruct PMF from other collective variables (e.g. Asp--amine and Trp--benzene distances). Is there a way to reweigh? Also how do I account for trajectories that were pruned (removed)? Is there an example?

Thank you,

Paween

[Daniel Zuckerman]

Paween, I can make some general comments here:

1. Remember that you won't get a true PMF from WE simulation unless you are using 'equilibrium' setting (no recycling at a target state) AND your WE has run long enough to relax to equilibrium ... which is highly unlikely for a complex system.
2. What you get from WE/WESTPA in general is the time evolving probability distribution which is relaxing toward either equilibrium or a non-equilibrium steady state (if recycling).  This is definitely worth examining for interesting features and a rough picture of convergence.
3. If you want to examine a coordinate besides what you used for bins, that is straightforward because you can just make a histogram for your coordinate(s) of interest using the weights (e.g., for a range of iterations).  If you take the negative log of this histogram, it will be free-energy-like ... though it is unlikely to be the equilibrium free energy as noted above.
4. Pruned trajectories are automatically accounted for in the weights, which is a nice thing about WE.  At any time, WE gives an unbiased estimate of the probability distribution as it evolves from the initial condition.  This is because WE resampling guarantees pruned weights are re-allocated in an unbiased way.
   
5. A separate subtlety is the issue of PMF vs free energy, in case you're not aware of it.  This has to do with Jacobians.  For example, if you have an equilibrium distribution that is uniform over a circle, but you plot a histogram in r and then take a log, you might think there is a (mean) force present, but there is not.  So ... be careful.
   

Bottom line: not only does WE have limitations in terms of sampling power (like any method) but we have to be a little more careful when analyzing WE data.  Well, I guess everyone needs to be more careful, but we shouldn't assume that a 'fancy' method always gives reliable or easy-to-interpret results.  On the other hand, examining alternative coordinates in WE is easy ... you just have to realize the meaning of the WE framework.  Take a look at the WE review article that Lillian Chong and I wrote a few years back.

--Dan Zuckerman

---

From: westpa...@googlegroups.com <westpa...@googlegroups.com> on behalf of Paween Mahinthichaichan <mahi...@gmail.com>
Sent: Monday, December 12, 2022 7:27 AM
To: westpa-users <westpa...@googlegroups.com>
Subject: [EXTERNAL] [westpa-users] reconstructing PMF from CVs other than progress coordinate

 

--
You received this message because you are subscribed to the Google Groups "westpa-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to westpa-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/westpa-users/2449f771-e09e-451e-a58a-2bd0d6cc25a6n%40googlegroups.com.

[Paween Mahinthichaichan]

Hi Dr. Zuckerman,

Thank you very much for the answers and the insights. These are thorough and helpful.

Paween

[Leung, Jeremy]

Hi Paween,

In addition to what Dr. Zuckerman has posted, here's advice on the technical side of things:

If you would like to plot the time-evolving probability distribution with your auxiliary data, you can follow the instructions here: https://github.com/westpa/westpa/wiki/man:w_pdist#analyzing-auxdata-datasets.

If you don't already have the data already saved during the simulation, you can follow the w_crawl tutorial here: https://github.com/westpa/westpa2_tutorials/blob/main/paper/AdvancedTutorialManuscript.pdf with examples here: https://github.com/westpa/westpa2_tutorials/tree/main/tutorial-3.1.

Hope this helps too!

Best,

Jeremy L.

--
Jeremy M. G. Leung
PhD Candidate, Chemistry
Graduate Student Researcher, Chemistry (Chong Lab)
University of Pittsburgh | 219 Parkman Avenue, Pittsburgh, PA 15260
jml...@pitt.edu | [He, Him, His]

To view this discussion on the web visit https://groups.google.com/d/msgid/westpa-users/44a45feb-046e-4e36-a5ea-c65a8c96d060n%40googlegroups.com.