[phenixbb] Disappearing side chains after autobuilding

[Tom Terwilliger]

Hi Katya,

Yes, this is a bug, and it is fixed in our working version, but you  
don't have it yet. I'm sorry about that! The problem was basically  
that autobuild tries to cross the original model (with truncated side  
chains) with a rebuilt model, and it uses the whole-residue  
correlation coefficient to decide which is better. If one of the  
residues has no side chain (i.e., if you supplied a sequence file that  
is different than the starting model) the one with no side chain  
doesn't include the side-chain region in the correlation. Sometimes  
that gives a higher correlation and the one with the side chain never  
gets taken. It should instead use the correlation over the same region  
for the 2 models...

All the best,
-Tom T



On Jan 21, 2009, at 2:43 PM, Katya Heldwein wrote:

> Hi,
>
> I ran phenix.autobuild with default setting (rebuild_in_place), and
> noticed that some side chains were gone in the resulting
> overall_best.pdb. That is, let's say, I started with Phe as residue  
> 10.
> After rebuilding, the Phe10 is still called Phe in the output file,  
> but
> the side chain atoms are gone, leaving residue 10 as Ala. This  
> happened
> for several side chains, and I am at a loss for why this happened. Is
> this some kind of a bug?
>
> Katya
>
> -- 
>
> Ekaterina Heldwein, Ph.D.
>
> Assistant Professor
>
> Department of Molecular Biology and Microbiology
>
> Tufts University School of Medicine
>
> 136 Harrison Ave
>
> Boston, MA 02111
>
>
>
> phone: 617-636-0858
>
> fax: 617-636-0337
>
> e-mail: katya.heldwein at tufts.edu
>
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://www.phenix-online.org/mailman/listinfo/phenixbb