Scott,
we can discuss this more offline if you want.
I'll post some comments here in case others are interested.
the bottom line is that these GB models (1&5)
have way too much alpha helix (Okur/Simmerling, JCTC 2006)
and GB1 has too much salt bridge (Geney/Simmerling JCTC 2006).
The structure distributions in GB and explicit water just don't match,
and these are all highly converged data. We have a few other
examples under review and a JMB paper in press.
I think that ff96 "looks" ok with GB because of cancellation
of error- ff96 doesn't like alpha but really GB does, and that's
dangerous since the GB part will depend on solvent exposure
while the backbone part doesn't. It's not at all transferable.
In addition ff96, 99 and a bunch of the other variants
have completely incorrect treatment of glycine.
we have a modified ff99 backbone parameter set that seems
to work much better, it's called ff99SB in amber 9 and the
manuscript is under review. I can send you a copy offline if
you're interested. We compare multiple backbone parameter
sets including those by Garcia and Pande.
Carlos
M. Scott Shell wrote:
> Carlos Simmerling wrote:
>
>> Scott,
>> I don't see any obvious problems but have you tried it without the
>> weight change?
>>
>> also unless this is just some test system I would really avoid using
>> ff96 as it is
>> quite bad, not just the (strong) beta bias but your glycine will
>> almost certainly be
>> very misbehaved. I would not use this combination and would have very
>> serious reservations about any article that I read where it was used.
>> We've also shown that igb1 and igb5 have a very strong alpha helical
>> bias, and
>> while the combination of gb and ff96 might sometimes seem to give
>> secondary structure
>> propensities in line with experiment it would be for all the wrong
>> reasons.
>> just my opinion... I really want people to stop using this parameter
>> set. Even the
>> Kollman lab where it was developed abandonded it very quickly.
>> Carlos
>>
>>
> I tried it without the weight change and without the DISANG line and
> the problem still occurs.
>
> I am very interested in your comments regarding ff96. We have
> performed a number of force field tests and found this one to be the
> best, for the time being, as a "middle of the road" parameter set for
> capturing both alpha and beta motifs in short peptides. Many of the
> newer sets, like ff99 and ff03, gave far-too-helical results for known
> stable hairpins. This seems in line with Okamoto's recent work (the
> Yoda et al paper in Chem Phys). Can you point me to some work on ff96
> showing its problems? Moreover, which implicit solvent/force field
> combination within Amber would you recommend? We are constantly
> searching for the best force field setup and are open to suggestions.
>
> -Scott
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sun Jun 11 2006 - 06:07:13 PDT
