[phenixbb] too many outliers

[Pavel Afonine]

  Hi Intekhab,

are all these outliers coming from that poorly modeled loop? If not (if 
yes - try to improve the loop first), here are a quick things to try:

1) try adding H and refine with optimize_wxc=true;
2) do "1)" after some geometry regularization;
3) try using secondary structure restraints;
4) try without SA;
5) fix all outliers manually in Coot and re-run the refinement. What 
happens then?
6) try manually some stricter weights (more impact on restraints) by 
using a number of smaller values of wxc_scale.

If nothing helps you can send me the data and model and I will have a 
look. Of course I don't have a magic stick and can't promise to fix 
it... Just in case I may spot something that may help...

Remember, a ramachandran or rotamer outlier is not necessarily wrong 
(example on page #22,25 here: 
http://www.phenix-online.org/presentations/latest/pavel_validation.pdf) 
but each of them must be explained and enough of evidence has to be 
provided to support your explanations (I'm speaking of final structure 
at publication stage).

Pavel.


On 8/13/10 12:44 AM, intekhab alam wrote:
> Hi there
> I have recently solved a structure at 2.71A resolution using CCP4 
> (Refmac5) with R and R free of 22 and 29 respectively. The model is 
> nicely fitted in the map and everything looks except in some loop 
> region where the map is not fitted well. I just finished modelling and 
> checked ramachandran plot which is showing unreasonably high number of 
> outliers (12%) with only 65% in the favoured region. I am thinking to 
> use phenix_refine with  simulated annaeling and using the option fix 
> bad chain rotamers. What else i can do aprt from this. i will be 
> highly thankful for the suggestions. Apart from this can anyone let me 
> know how
>
> -- 
> INTEKHAB ALAM
> LABORATORY OF STRUCTURAL BIOINFORMATICS
> KOREA UNIVERSITY, SEOUL