Analysis of AWH simulations

User discussions
,
1

GROMACS version: 2024.1
GROMACS modification: Yes/No
Dear all,

I just learned about the AWH method, and found it very interesting and powerful (I am simulating the binding of small ligands into protein active sites). I read a lot of tutorials, papers and most of the posts here which helped to optimize my simulation parameters, but I still have some specific doubts.

As an initial test case, after equilibration steps I ran a a multi-walker awh simulation with 2 walkers (t = 100 ns) and observed the following in the log files:

  1. equilibrated histogram at t = 28927.8 ps
  2. awh1: covering at t = 36003 ps. Decreased the update size; awh1: out of the initial stage at t = 36003.

Question A:
If I wish to perform other analysis (such as extract relevant frames close to each PMF minima), those should be performed considering the simulation after ~36 ns? Since there are two trajectories, could I skip the initial 36 ns of each and concatenate them for further structural analysis?

Question B:
although these 100 ns test simulations are very short (most papers I read simulated more than 500 ns for each walker), PMF profiles between 90-100 ns seemed “converged”, with differences below 0.2 kT. Is this a sufficient criteria?

Question C:
To streghten tha results, I plan to run two more “replicas” for each protein-ligand system, in which the initial structures are different (different protein conformation and ligand poses) prior to minimization/equilibration steps and mw-awh simulations. Is this OK?

Kind regards,
Felippe

2

A: Yes. But note that you will be analyzing a biased ensemble, not a Boltzmann ensemble.
B: Yes, if you run a sufficient number of simulations, at least 3 or 4.
C: Yes.

3

Thanks, Hess!