Re: [phenixbb] question on bad geometry
Hi Fengyun, Just to emphasize what Ed points out here, an essential step in completing a structure is to compare it, one residue, solvent molecule, or ligand at a time, with your electron density map, fixing any inconsistencies between the model and the map. There are many wonderful tools in all the crystallographic graphics programs to help you do this. All the best, Tom T
On Tue, 2010-08-31 at 20:37 -0500, [email protected] wrote:
Does that mean I need to manually adjust the model?
Autobuilder is a fantastic tool which greatly simplifies the model building process, but you always must validate your structure by inspecting the electron density. I fully understand that (hopefully) it is what you plan to do and this is more of a reaction to the alarming number of obvious and easily correctable systematic errors one finds in structures deposited in the PDB these days.
Thanks to Pavel for providing an excellent example to illustrate the point. One only needs to take a look at the model in context of electron density maps to confirm that autobuilder will fix some errors but also sometimes introduce some that require manual correction.
Cheers,
Ed.
_______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb
Hello, On 9/1/10 5:25 AM, Thomas C. Terwilliger wrote:
Just to emphasize what Ed points out here, an essential step in completing a structure is to compare it, one residue, solvent molecule, or ligand at a time, with your electron density map, fixing any inconsistencies between the model and the map.
there are tools to help with this. "Comprehensive validation" available in PHENIX GUI will give you the map correlation computed for each residue (or even atom), so you can pinpoint problem residues in no time just by sorting the list and displaying only those having map CC less than say 0.8. Combined with the list of geometry outliers (Ramachandran, bad clashes, rotamer and Cb-deviation outliers - this is all reported by "Comprehensive validation") and the functionality to click on the outlier and immediately get it zoomed in Coot, makes addressing the problems much easier. Pavel.
Thanks for all the suggestions. I am now checking the model manually. With the "rotamer analysis" in coot, the plot shows that most of the rotamer outliers are Leu. Is this the common case that Leu appears as the rotamer outlier most often? In my model, there is a 20-residue loop region (not metal binding). The geometry for this region is bad and also there is no obvious density in 2FOFC-map. The density for this loop only shows up after density modification in dm. I'll first try to fix the other parts with clear density. Thanks everyone for the reply! Regards, Fengyun
Hi Fengyun,
On Wed, Sep 1, 2010 at 11:12 AM,
I am now checking the model manually. With the "rotamer analysis" in coot, the plot shows that most of the rotamer outliers are Leu. Is this the common case that Leu appears as the rotamer outlier most often?
Yes, Leu sidechains are often misfit. It's very easy at medium resolution to fit a Leu sidechain ~180 degrees backward, which is almost always an unfavorable conformation. Flipping them back around should arrive at a highly populated, favorable conformation. This problem also effects Thr and Val sidechains, and Ile at times, too. You can read more about this here: http://www.ncbi.nlm.nih.gov/pubmed/19002604 Coot is very good at fixing these errors with the 'autofit best rotamer' tool, and most should also be fixed with the PHENIX 'fix rotamers' tool. Jeff
participants (4)
-
fn1＠rice.edu -
Jeff Headd -
Pavel Afonine -
Thomas C. Terwilliger