The problem of all these metrics is that they are binary: YES (=outlier) or NO (=not an outlier).

You may be 1.e-5 or so (just making up a number) away from NO but the program may still consider it as outlier. And obviously you will not see that different on graphics (in Coot for instance).

This is exactly why I don't particularly like functions with kinks and instead prefere smoothly differentiable ones!

Phenix.refine uses CB restraints by default, so unless you disabled it or your model is very poor, you should not have these outliers after refinement (again, normally).

Pavel

On 11/5/14 11:43 AM, George Devaniranjan wrote:

Hi,

I refined a structure using phenix.refine.

The R/R free decreased slightly from 13.46/15.87 to 13.0/15.6.

But, molprobity said there were 3 CB outliers (starting from 0).

I looked at those residues in COOT. I could NOT really see a different between the refined and initial model.

In addition I looked at COOT

validate-->density fit analysis

I cannot see ANY difference in the peak heights (same numerical value too).

I looked at RSCC (using phenix validate tools as well).

I still can't see a difference in the plots.

I am not sure how to get rid of the 3 CB outliers that have crept in since I can't pick them out as worse off in the refined model.

I ask since in another case > 3 CB outliers, with the same results in picking them out in COOT/RSCC.

I am wondering if there is another way I can see the deviation and fix it?

Thank you,

George
_______________________________________________
phenixbb mailing list
[email protected]
http://phenix-online.org/mailman/listinfo/phenixbb