protien crystalography.

Kaj Stenberg kstenber at cc.helsinki.fi
Sat Nov 8 16:48:28 EST 1997


(I think that bionet.xtallography would be a better place for this 
question)


WWolfgong <wwolfgong at aol.com> wrote:

>I am a staff crystalogapher for a chemistry department. This involves 
>solving structures for chemists (molecules which rarely exceed 200 
>hundred nonhydrogen atoms ( this would be the extreme case)). My
>question is how does protien crystallography differ. I understand 
>the resolution is lower (~1.5 angstroms versus ~0.8) and that 
>instead of assigning atoms to the difference map, amino acid residues 
>asigned instead during refinements of the model. 

Below 2 A is actually high resolution for a protein structure

>How do you chose a residue and what is the justification? 

Usually the gene coding for the protein is known, so the string of 
amino acids is clear. Mind you, there are sometimes errors in the gene 
sequence, in these cases one hopes to have a clear enough electron 
density map to assign the right amino acid.

>For me, I look at bond lengths and angles, how could that apply to 
>amino acid residues, I see no relation.

Bond lengths and angles within amino acids are generally idealized 
and taken from the crystal structures of the sole amino acid. The 
side chains have a limited number of favoured rotamers, here the 
need for manual interpretation (between refinement cycles) of the 
map versus model comes into play. 
 
>Also, a good structure for a compound has a R value of less than 6%, 
>I have seen publications of protien structures which exceed 30%. What 
>is good for protiens?

Rule-of-thumb (conventional R):  good <20% acceptable < 25% 
However, the resolution is of importance, as well as the stereochemical 
"sense" of the model. An unwanted powerful feature of refinement 
programs is that the (conventional) R-factor can be lowered by testing 
the parameters, and the structure gets "overrefined", that is refined 
beyond the support of experimental data. To overcome this several other 
R-factors have been developed, most noteworthy the FreeR-factor. Here a 
portion of the data is taken aside and newer used in the refinement, 
only in the R-factor calculations, this to avoid bias. FreeR is 
generally higher (a few-5%) than (conv.)R, but more reliable. Anything 
under 30% is acceptable (if the numerous other quality indicators give 
green light).

 
============================================================================
Kaj Stenberg, Ph. D 
Department of Biosciences            tel. +358-9-708 59032
Division of Biochemistry             fax +358-9-708 59068
P. O. Box 56, Viikinkaari 5          e-mail kstenber at kruuna.helsinki.fi
FIN-00014 University of Helsinki     http://www.helsinki.fi/~kstenber/
Finland

This message is enviromentally friendly: every bit was created using 
100% recycled electrons.

=============================================================================




More information about the Comp-bio mailing list