Reweighting metadynamics ver. 2.2 vs ver. 2.3

843 views
Skip to first unread message

Bogdan Marekha

unread,
Sep 19, 2016, 12:28:16 PM9/19/16
to PLUMED users
Hi folks!

Acoording to the changelog the reweighting algorithm of P. Tiwary and M. Parrinello.(JPC B, 119(3):736–742) has been implemented in the PLUMED since the version 2.2. However, the proper usage seems to have changed in the version 2.3.

Namely, in PLUMED 2.3 one has to specify the reweighting grid and frequency in the relevant metadynamics action (options REWEIGHTING_NGRID= and  REWEIGHTING_NHILLS=, respectively), then declare the corresponding reweighted metadynamics bias using the REWEIGHT_METAD action, and, finally, use the latter as an argument of the LOGWEIGHTS option in the HISTOGRAM action. It is also stressed that it is crucial to use this dynamically updated reweighted bias and not the static one from the REWEIGHT_BIAS action.

In other words, if one wishes to get a 2D free-energy map as a functions of 2 CVS having performed the metadynamics only using CV1 the key commands would be:

METAD ...
ARG=CV1
... # metadynamics relevant options
REWEIGHTING_NGRID=100 # GE than the original number of grids of this CV
REWEIGHTING_NHILLS=50
... METAD

mdbias: REWEIGHT_METAD TEMP=300

HISTOGRAM ...
ARG=CV1,CV2
LOGWEIGHTS=mdbias
... # histogram relevant options
... HISTOGRAM

And finally use the CONVERT_TO_FES utility.

Conversely, in PLUMED 2.2 there is no REWEIGHT_METAD action, only the REWEIGHT_BIAS option of the HISTOGRAM action, which takes all the available biases into account for the reweighting. If one wishes to perform the same task, the relevant commands would be

METAD ...
ARG=CV1
... # metadynamics relevant options
REWEIGHTING_NGRID=100 # GE than the original number of grids of this CV
REWEIGHTING_NHILLS=50
... METAD
HISTOGRAM ...
ARG=CV1,CV2
REWEIGHT_BIAS
FREE-ENERGY
TEMP=300
... # histogram relevant options
... HISTOGRAM

I have two questions here:

1. Does it mean that the version 2.2 implementation of the cited reweighting algorithm is not correct as it takes the static (not dynamically reweighted) metadynamics bias and potentially other biases, e.g., from restraints?
2. What kind of reweighting is done in version 2.2 if one uses the REWEIGHT_BIAS option in the HISTOGRAM action, but does not specify the REWEIGHTING_NGRID and  REWEIGHTING_NHILLS options in the METAD action? Is it the one from Bonomi, Barducci, and Parinello (J. Comp. Chem., 30(11),:1615-1621?

Cheers,
Bogdan


Gareth Tribello

unread,
Sep 19, 2016, 1:45:49 PM9/19/16
to plumed...@googlegroups.com
Hello Bogdan

Thanks for your email and for pointing this out.  If you look in the change log for version 2.3 it says that

  • The way that reweighting is performed is now different. There are three separate actions REWEIGHT_BIASREWEIGHT_TEMP and REWEIGHT_METAD. These actions calculate the quantities that were calculated using the keywords REWEIGHT_BIAS and REWEIGHT_TEMP that used to appear in the old HISTOGRAM method. Now those these methods can be used in any methods that calculate ensemble averages for example HISTOGRAM and AVERAGE
To be clear though we are always using the P. Tiwary and M. Parrinello algorithm (in version 2.2 and version 2.3).  There was nothing wrong with the version of the algorithm in 2.2.  The versions in 2.3 and 2.2 are both correct.  This change is just a change in syntax so that we can use the same reweighting code for calculating histograms and all ensemble averages.

Thanks again
Gareth

--
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 post to this group, send email to plumed...@googlegroups.com.
Visit this group at https://groups.google.com/group/plumed-users.
To view this discussion on the web visit https://groups.google.com/d/msgid/plumed-users/33c213f8-ac00-4972-80e4-07107d17109f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Bogdan Marekha

unread,
Sep 20, 2016, 11:24:24 AM9/20/16
to PLUMED users
Hi Gareth,

thanks for the clarification that the correction hasn't essentially changed.

I am wondering though how  the periodic reweighting of the bias (invoked by the REWEIGHTING_NHILLS and REWEIGHTING_NGRID options) influences the classical metadynamics analysis using e.g. sum_hills to reconstruct the FES as a function of the biased variables. Should we use it only when doing the HISTOGRAM reweighting?

Best,
Bogdan

Gareth Tribello

unread,
Sep 20, 2016, 11:35:42 AM9/20/16
to plumed...@googlegroups.com
Hello

I am not sure what you are asking here.  You can calculate the free energy by summing the hills and by reweighting.  You shouldn’t have any problems here as by turning on the REWEIGHT_METAD command you are not doing anything to modify what is output in the HILLS file.  Similarly for the flag that you have to add to the METAD action.  What is output to the hills file with and without that flag will be the same

Gareth

Reply all
Reply to author
Forward
0 new messages