usage of abf_integrate

[meghaparashar2]

Dear all,
I am using eABF free energy method to calculate dimerization energy. For this I have divided the 1d CV space in different windows and did the simulations. Now I am having .czar.grad files as well .hist.grad files. I used InputPrefix to get the PMF. Now to check the convergence I want to use abf_integrate but documentation is not very clear on how to use it. I had few questions about it, like:

1. Can I supply .hist.grad files as I want to see the time evolution of RMSD of PMF I got from eABF and from abf_integrate, if not, then how can that be done?
2. How one can get the evolution of PMF as a function of simulation time using, can it be done using .hist. files?

Thanks,
Megha

[jhenin]

Dear Megha,
This is a fair question, the workflow for this is a little complicated.
abf_integrate will only integrate a single gradient grid, it won't integrate all the time steps stored in a history file. The good news is that history files are just concatenated grid files, so you can split the hist.grad file and process each of the fragments with abf_integrate then concatenate the resulting free energy files.
Once you have a history file, the simplest way to look at convergence is with this script:

import colvars_grid as cvg
import matplotlib.pyplot as plt

pmf_traj = cvg.colvars_grid('mysim.hist.pmf')
plt.plot(pmf_traj.convergence())

This uses the Python module found here:
https://github.com/Colvars/colvars/blob/master/colvartools/colvars_grid.py

[meghaparashar2]

Dear J Henin,
Thanks for your quick reply.
While going through the paper I noticed there are 4 convergence criterion used by the authors (Fig.5 & Fig.6).
But the colvars_grid module is acting like a black box, I have no idea what is going on.
I had few questions about this like:

1. How the RMSD of PMF is being calculated, what is the by default user provided numerical value (if it is working like abf_integrate)?
2. How can I calculate Sampling entropy, Sampling heterogeneity?
3. I am assuming RMSD of grad can be obtained in a similar fashion as thethe RMSD of PMF was obtained, am i right?
4. How the error bars were calculated in the paper?

Can you please refer me to a more explained documentation of these?

Thanks much!
Megha

[meghaparashar2]

Dear J Henin,
I think I have understood most of the part of convergence calculations.
Though I have to ask one thing, what does 'nats' mean, that's you have used in the plot of sampling entropy and sampling heterogeneity?

Thank you!
Megha

[jhenin]

Nats are the natural unit of entropy: https://en.wikipedia.org/wiki/Nat_(unit).