basics for WellTempered metadynamics

[Emmanuel Autieri]

Hi everybody,
I need to address simulation with sugars using the well-tempered metadynamics.

I have a (maybe very simple) question about postprocessing in this case. It's quite clear to me how to handle data analisys (fes evaluation, errors estimation, etc... ) with standard metadynamics, but it is not the same with the well-tempered version.
Can someone help me (with indications, references, and so on)?

Thanks in advance
E.A.

--
Emmanuel Autieri
PhD student
    Interdisciplinary Laboratory for Computational Science (LISC)
    Via Sommarive, 18 - 38123  Povo (TN) - Italy
tel. +39 0461 314651   cell. +39 347 7175989
e-mail aut...@science.unitn.it
          emmanuel...@gmail.com
          aut...@fbk.eu

[Giovanni Bussi]

Dear Emmanuel,

WT-metaD is very similar to plain metaD, just take into account the
following points:
* hills size is decreasing along the simulation, so that bias is going
to *converge* (and not *oscillate*) in the long run
* final bias is not completely canceling the underlying fes, so that
exploration is not uniform. in particular, more time is spent on
minima than on maxima of the fes, according to an apparent "high
temperature" distribution of the collective variable (see BIASFACTOR
option).
* the HILLS file does not contain the bias but already the estimate
for the fes. thus, you can use the sum_hills tool as usual
* concerning error, fes estimate from WT-metaD is similar to a
time-averaging. thus, plot the final result (e.g. the free-energy
difference between two minima) as a function of time, and it should
converge to its exact value (a part from, of course, force-field
errors).
* also here, as in plain metaD, you need long simulations and good
collective variables to hope in a convergence!

Giovanni

> --
> You received this message because you are subscribed to the Google Groups
> "PLUMED users" group.
> To post to this group, send email to plumed...@googlegroups.com.
> To unsubscribe from this group, send email to
> plumed-users...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/plumed-users?hl=en.
>

[Andrea Spitaleri]

Ciao Emmanuel,
well I have been working with well-tempered since a while. I will add
few more points to what Giovanni said.

i) to check the convergence in well-tempered I use to monitor the
hills height vs time (or the bias). at convergence you should see as
stated by Giovanni the hills profile going to 0. however, this is not
always true.. Sometimes I do not see any jump in the height for a
while and after waiting for a long time another minimum is found. This
in my case was indicating a neglected CV.

ii) to estimate error in the FES, I use to run 3-5 well-tempered
simulations and average the profiles from them.

iii) for my cyclopeptides (5 aa) I used T = 1000K to study the FES in
the phi/psi dihedral angle space. Increasing that factor, the FES
profile was getting "flatter" (the saddle point were going down).
Probably since the system is small, I was puting too much energy.

iv) you may use the reweighting algoritm (seen in the workshop) to
recalculate the FES in another space using different CV

hope it helps

and

2011/3/14 Giovanni Bussi <giovann...@gmail.com>:

--
----------------------------------------------------------------
http://sites.google.com/site/andreaspitaleri/
----------------------------------------------------------------
http://it.linkedin.com/in/andreaspitaleri
----------------------------------------------------------------