Re: [AMBER-Developers] Amber atom type naming (and parms)

From: Yong Duan <>
Date: Wed, 14 Dec 2011 16:41:02 -0800

I think 109.6 is better.

I actually think we should use standard bond angles, 180, 120, 109.5, 107
and avoid (over)fitting to any specific molecules.

We recently ran into a funny case in which the nucleic acid base can be
notably deformed to almost tetrahedral geometry. It's actually not a new
case. I think it has been there since the beginning of amber nucleic acid
force field. If you look at parm91.dat, a lot of bond angles are notably
off the canonical values. Magically, when you put them together, they
typically make sense. In other words, these parameters were designed just
to make sure those particular molecules (nucleic acids and amino acids)
were well-modeled. However, in special cases, they could just become
funny. For example when a planar sp2 carbon's three bond angles sum
together is notably smaller than 360, you'd have the base no longer
planar. This is not a new issue. Those of us who have been around long
enough know this all along. For long time, the solution was to use
improper to compensate such defect. The problem is that bond angle
energies are typically at 50 kcal/mole/rad**2. You'd need improper
parameters of comparable magnitude to compensate that, just plain
ridiculous. The root cause of these problems, IMHO, is the over-fitting of
the bond angles to a particular (set of) chemistry. I suspect that the
<OS-P-OS = 102.6 was also such case. Fortunately, (or unfortunately), it
is not part of the planar construct. Therefore, nobody has paid much
attention to it.

Glad you guys pick up this.

Yong Duan, Ph.D, Professor
UC Davis Genome Center and
Department of Biomedical Engineering
University of California at Davis
Davis, CA 95616
On 12/14/11 3:23 PM, "B. Lachele Foley" <> wrote:
>Sounds ok.  I worry that fragment-based parms could introduce a lot more
>complexity and associated entropic drift to have to contain.  Otherwise,
>yeah, cool.
>I've been tooling around with the Glycam param set to make it as
>orthogonal to parm99 and parm10 -- and everything else -- as possible.  I
>still need to test more, but I think I managed to make it so the load
>order mostly doesn't matter in most cases.
>On that subject, I could use input:
>* A conflict is introduced in parm10 -- the type CP (also long unused) is
>being used again there.  I just learned that today.  I'm arguing that we
>should give in on this one.  Not sure if I'll win.
>* We changed the params in angle OS-P-OS to a value that better fits
>crystal structure data for nucleic acids, lipids and carbs.  There is a
>conflicting (and older) parm set in 94/99/10.  Any chance y'all might
>want to adopt ours?
>Older one:          OS-P -OS     45.0     102.60   Parm94
>Ours:                 OS-P -OS     45.0     109.6    Parm94 and changed
>eqm value for better reproduction of average of crystal structures
>:-) Lachele
>Dr. B. Lachele Foley
>Complex Carbohydrate Research Center
>The University of Georgia
>Athens, GA USA
AMBER-Developers mailing list
Received on Wed Dec 14 2011 - 17:00:03 PST
Custom Search