[phenixbb] Bad ligand parameters

Joe Krahn krahn at niehs.nih.gov
Mon Sep 28 16:18:10 PDT 2009

GeoStd seems not to have an files. Is it only available via SVN for now?

The GeoStd parameters still need some work. The equivalent hydroxyls are 
not symmetric. One of the dihedrals has a base angle of -175.0 instead 
of 180 (or +/-60). One has an esd of 30.0, and the others have an esd of 
20.0. The hydrogen ESDs also seem a bit too tight, and probably should 
have a periodicity of 3.

They are an improvement, but I don't think automated parameterization 
will ever be perfect.


Nigel W Moriarty wrote:
> Joe
> It is in the nightly builds if you are inclined.  You are also free to 
> download the restraints from the GeoStd open source project.
> Nigel
> On 9/28/09 1:43 PM, Joe Krahn wrote:
>> Nigel,
>> I am using PHENIX 1.4-6, which is newer that the most recent version on 
>> the PHENIX web site, other than "nightly build" versions.
>> Nigel W Moriarty wrote:
>>> Joe
>>> I think you may need to upgrade your version of PHENIX.  In the current 
>>> version, the periodicity is 3.  I have recently been using eLBOW, the 
>>> Chemical Components library and the Monomer Library to improve any 
>>> residue restraints that have short comings.  These are freely available 
>>> at Source Forge from a project called GeoStd at 
>>> http://sourceforge.net/projects/geostd
>>> Its included in more recent PHENIX versions in $PHENIX/chem_data/geostd
>>> Nigel
>>> On 9/28/09 11:45 AM, Joe Krahn wrote:
>>>> I found that residue TRS has different dihedral target angles for the 
>>>> three hydroxyl oxygens, which are actually all equivalent. They have a 
>>>> periodicity of 1, but should have a periodicity of 3. What is the best 
>>>> way to handle such discrepancies?
>>>> Most ligand parameters are probably auto-generated based on a reference 
>>>> structure, and there is probably a lot of conformation bias. In this 
>>>> case, it seems that parameterization should have been able to detect 
>>>> symmetry. It would be nice if parameters could include information about 
>>>> whether a human has done any validation. Actually, a ligand Wiki might 
>>>> be nice, where people can easily put comments, even if they are not sure 
>>>> about how to improve the parameters.
>>>> Joe Krahn

More information about the phenixbb mailing list