AMBER Archive (2007)

Subject: Re: AMBER: Harmonic constraints

From: Carlos Simmerling (carlos.simmerling_at_gmail.com)
Date: Fri Sep 07 2007 - 14:06:35 CDT


I completely agree with Tom - I had just meant that for the
kind of simulations most people are doing, IMO the community
should no longer accept ff99-based data (for proteins)
without extremely compelling evidence that the results
are reliable. Of course there are many kinds of simulations that
may be ok with ff99 (such as very short or completely restrained runs),
and I don't mean to imply that old data with ff99 was always wrong.
My point is just that we know there are serious limitations
and there are better options available that do not increase simulation
complexity, so we as a community should expect new research
to either use the better models or else explain why it was not done.
It's not unlike using a distance-dependent dielectric treatment for
water- it might still be useful in some cases, but generally I think
we all place higher demands for validation of that kind of data
than we might with a "better" treatment of solvation. Ideally all
simulations should be validated against experiment, but we
should be more suspicious of data generated using models that
have serious and well-documented weaknesses.

I'm stressing this point because I find many people remain unaware
of the improvements in force fields over the past years and still
begin new projects with old/poor parameters. Of course ongoing projects
may require old parameters for consistency, but one should be
aware of the limitations (and as Tom says, address them accordingly
in publications). Unfortunately not all reviewers are requiring this.
carlos

On 9/7/07, Thomas Cheatham <tec3_at_utah.edu> wrote:
>
>
>
> Finally, if force field artifacts or limitations are known, it becomes
> incumbant on the user or paper author to acknowledge those limitations,
> note how they may influence their results, and convince the reviewers or
> readers that these artifacts do not unduly influence the results. This
> gets back to the assessment and validation of the results; we can model
> anything, but do we actually model reality and/or explain experiment?
>
> -- tec3
>
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu