AMBER Archive (2009)

Subject: Re: [AMBER] ff03 parameter for GDP and GTP

From: FyD (fyd_at_q4md-forcefieldtools.org)
Date: Sun Oct 11 2009 - 06:42:25 CDT

Dear Carlos,

> that's excellent that you have these available. I'm looking for similar
> parameters (but not ff03). Can you post the reference to the article?

Did you do a biblio. search ?

This work (& others) were presented to the 236th ACS National Meeting:
C. Cezard, E. Vanquelef, J. Pecher, P. Sonnet, P. Cieplak, E. Derat &
F.-Y. Dupradeau, RESP charge derivation and force field topology
database generation for complex bio-molecular systems and analogs,
236th ACS National Meeting, Philadelphia, PA, USA, August 17 - August
21, 2008.

As an ACS organizer, I guess you know this reference & it is also
available from the q4md-fft web site.

> thanks- I did know that one, I was just wondering if Francois had also
> published something so I could see how the validation was done.

First, I would like to underline that charge values generated by the
R.E.D. tools are reproducible. This means that every user can
reproduce the data generated. The last III.3 distribution present a
new script allowing to directly demonstrate this reproducibility using
Gaussian, PC-GAMESS, GAMESS-US & WinGAMESS under UNIX, Windows &
MacOS. Obviously, reproducible does not mean suitable for MD
simulation, but at least it provides a starting point to characterize
and then to improve a force field library. It also shows that charge
derivation was correctly or not correctly done by the user & this is
also convenient for new users.

Then, a R.E.DD.B. project is obviously not a publication, but the
computational conditions described in a R.E.DD.B. project are often
more detailed than these in the corresponding paper since R.E.DD.B.
does not only store text but also RESP inputs, Cartesian coordinates
of optimized geometries (& even more) besides the FF libraries
themself. Thus, we think a R.E.DD.B. project is complementary to a
publication.

Finally, R.E.DD.B. presents now a peer-review system for each project
submitted before release. Thus, the data we submit in the database we
have developped have also to follow this peer-review system.
Researchers from the CHARMM, AMBER, GLYCAM, PC-GAMESS and GROMACS
communities agreed to participate. Please, read the R.E.DD.B. FAQ
about this peer-review system. As expected it helps a lot to improve
the projects submitted & the database itself.

> that's good news,

Nothing new here. All R.E.DD.B. projects follow this principle - not
only this co-factor FFTopDB. Please, read the R.E.DD.B. FAQ.

> and I strongly urge you to also publish the parameters in
> the peer-reviewed literature so that less experienced users can be confident
> about adopting them even if they aren't able to evaluate the conditions
> themselves.

"Urge" me to publish our data ? ... Thanks a lot for this advice.
Well ... like others we are working on it ;-)

However, you are right: at least our data should have been released in
R.E.DD.B. so that every user can study what we have done. This has to
be done ASAP since we have distributed some of these FF libraries to
users. In fact, it turns out after that ACS meeting that R.E.DD.B. had
to be modified to accept such a big force field topology database
(FFTopDB). Now R.E.DD.B. is fully functional for such a big project,
but we never took the time to submit them. This is bad. OK let's do it.

Concerning the validation, I agree with you & validation is not always
easy for such molecular systems. Characteristic examples of this
FFTopDB were used and are currently used in MD simulations in our
groups.

> In my opinion it's very important for the validation aspect of
> the any group's parameters to be reviewed independently.

Once again I agree with you. However, as I said R.E.DD.B. presents
such an independent peer-review system and as I said R.E.DD.B. does
not only store sentences... Thus, we do beleive that R.E.DD.B. could
be a first step before publication when new FF libraries with RESP and
ESP charges are generated.

> It's not so much
> about comparing raw numerical values for the parameters like you mention
> below, but seeing how well they perform in use.

I do not understand the reason of this "but", raw numerical values can
tell a lot, Yes; charge values can be compared AND charge values have
to be validated to see how they performed.

Would you accept to participate in the R.E.DD.B. peer-review system ?
You would help us a lot by your great expertise. Would you be
interested to be one of the reviewer for this co-factor project before
its official release in R.E.DD.B. ? Obviously, as we decide to ask for
a review for this work and we handle this database this review cannot
be anonymous. But if every researcher follow what is described in the
FAQ, I do not think this is a problem. However, I am sure you are
involved in more important projects.

I hope I answered to all the points you have underlined.

regards, Francois

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber