AMBER Archive (2006)

Subject: AMBER: ff03 and ff99SB force fields

From: Carlos Simmerling (
Date: Thu Aug 03 2006 - 05:59:11 CDT

I'm replying off the list to send you the manuscript
for the ff99SB article that is in press at Proteins.
For the purposes of the archive, I'll clarify my comments on
compatibility and invite Yong Duan to join in and give us the
benefit of his much more extensive experience with ff03.

I think that one has to be extremely careful mixing force fields.
To my knowledge it has not been demonstrated that hydrogen
bonds (for example) between ff94 and ff03 backbone charges
are the same as ones between pairs of ff03 groups. Since the charge
model fundamentally differs (or else why is there a need for ff03?) one
should probably not mix them in a single simulation. It would be like
charmm parameters and using with ff94- both are internally consistent but
you need to carefully test whether they have the same intra- and inter-
force field interactions.This is important to me since one of my projects
involves protein/DNA complexes and I don't want to do all of the
extensive testing needed to make sure that the DNA parameters will
work well with ff03 proteins when I know that ff99SB doesn't have
that concern.Another concern about ff03 is that not all of the
torsions were refit. I don't mean to criticize Duan's effort at all,
just to
point out that I think ff03 is still a work in progress (and a good one).

Our ff99SB paper is in press at Proteins. I believe that we do as well
or even
slightly better than ff03 for the things you ask about. Both ff99SB and
ff03 are
much better than any of the other Amber variants that we tested. In addition
to secondary structure bias, almost all of the other have very serious
problems with their glycine parameters.

Vlad Cojocaru wrote:

>Dear Carlos,
>I am following closely this discussion about the incompatibilities
>between ff03 and the other amber force field versions. So, first I would
>like to ask you, what do you actually mean by "imbalance in elecrostatics".
>Furthermore, the paper describing the ff03 speaks of correlation between
>the "new" charges and the old charges and claims that at least in cases
>of ligand-protein interactions, there should be no reason not to use
>"old" ligand charges with "new" charges for protein residues.
>Moreover, when ff03 is loaded into Leap, only non-terminal amino acids
>are loaded from the new library, while N-terminal and C-terminal AA are
>loaded from the old 94 libraries. So, isnt that inconssistent? Or is it
>believed that non-terminal residues are not that important?...
>The reason for asking these questions, is because according to studies
>that have been done in our group (not by me), ff03 seems to be by far
>the best version if one is interested in lifetimes and stabilities of
>secondary structural elements (ff99SB was not tested). Is there any
>reference which compares ff99SB with ff03 in that respect?
>Best wishes
>Carlos Simmerling wrote:
>>due to the different charge model, you will need to find parameters
>>developed specifically for ff03. using other sets would result in
>>imbalance in electrostatics between different groups and is not
>>a good idea. On the other hand, ff94, ff96, ff99 and several other
>>variants such as our lab's ff99SB and parameters from Garcia and
>>Pande all use the same charge model and will work with any parameters
>>developed for ff94.
>>Zu Thur Yew wrote:
>>>I would like ask if there are any pTHR parameters suitable for use
>>>with ff03?
>>>I know there are parms developed in the same way as ff94/ff99...the
>>>charges for normal THR in ff99 & ff03 seems quite different...
>>>has anyone any experience using these parms with ff03? Would problems
>>>be expected if these parameters are used with ff03?
>>>Any advice would be most appreciated!
>>>Many Thanks,
>>>The AMBER Mail Reflector
>>>To post, send mail to
>>>To unsubscribe, send "unsubscribe amber" to
>>The AMBER Mail Reflector
>>To post, send mail to
>>To unsubscribe, send "unsubscribe amber" to
The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to