reweighting with OPES

[Andrey Poletayev]

Hello folks, 

I have what is likely a dumb question regarding the reweighting procedure for OPES. 

For well-tempered metadynamics, we can reweight a trajectory using REWEIGHT_METAD, or chop off some starting region where the bias is changing rapidly and use REWEIGHT_BIAS. If we use the REWEIGHT_METAD, we can also obtain thermodynamically weighted quantities from the same trajectory as a function of the CVs. For example, if I am tracking box dimensions across a phase transition driven by the CV, I can get all of that from REWEIGHT_METAD and histograms. 

I would like to find an equivalent for OPES: thermodynamic averages from an OPES-biased simulation.

If I understand correctly, the reweighting suggested in tutorials (link) is equivalent to REWEIGHT_BIAS, which would not work for this purpose. It is also not recommended to use the OPES_METAD's rct output for reweighting. Is there an equivalent of REWEIGHT_METAD for OPES - or a better way of obtaining averages over CVs from OPES trajectories? 

(If I'm missing something silly or something done in literature, I would appreciate pointers to literature or plumed-nest)

Andrey

[Michele Invernizzi]

Hi Andrey,
See this https://groups.google.com/g/plumed-users/c/uoi8S8m1T0s about the difference between REWEIGHT_BIAS and REWEIGHT_METAD.

You can get averages over any CVs (with OPES or METAD) with the same procedure that you use for CVs used during biasing.

Personally, I prefer to use python scripts for post-processing, see e.g. this tutorial https://github.com/invemichele/masterclass-22-03

Best,
Michele

--
You received this message because you are subscribed to the Google Groups "PLUMED users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plumed-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/03301770-90e9-4f43-8218-a4fc7860c043n%40googlegroups.com.

[alx.th...@gmail.com]

Hi Andrey,

REWEIGHT_BIAS and REWEIGHT_METAD do the same. With REWEIGHT_BIAS you have more control since you can feed any biases inside. For example, this way you can integrate out the effect of the walls you impose on your CV.

Initial transient should be cut out irrespective of which function is used. Same is true for OPES or OPES_METAD, though this period will be much smaller. OPES_METAD's rct should not be used for reweighting, since its bias is already reweight-ready. So, OPES bias == METAD rbias (bias - rct) in practical terms. You need OPES rct to check what is the time that you should discard (in OPES, rct should become constant to singify the time from which onwards you can start reweighting). In METAD rct should become roughly linear for the same purposes.

Doing some basic stuff greatly helped me to understand that. I.e. simulate some symmetric process, and then produce profiles with REWEIGHT while gradually increasing the initial transient time. If you do not discard at all, you will observe an overestimation of the initial state irrespective of how long you simulate. Just because your opes biases/metad rbiases for starting time frames are completely off the roof.

Best,

Alex