Ok, in this case I need to look at the files ... Pavel On 4/19/12 5:46 AM, Mario Sanches wrote:

I am using a nightly build from around two weeks ago (dev-1021). I am sending you some files. Thank you,

Mario Sanches

On Thu, Apr 19, 2012 at 8:42 AM, Pavel Afonine mailto:[email protected]> wrote:

Hi Mario,

are you using the latest Phenix? If not, please update your version of Phenix and try again:

http://www.phenix-online.org/download/nightly_builds.cgi

If this doesn't help, please send me the data and model (offlist) and I will have a look to see what's going on.

Pavel

On 4/19/12 5:37 AM, Mario Sanches wrote:

...

I am refining a low resolution structure (2.9A) using phenix.refine. As I continue the refinement Rwork/Rfree are dropping but my geometry statistics in general are very bad. My clashscore is getting worst and the RMS(angle) is really high. I have tried to play with the "wxc_scale" parameter and also ran phenix selecting the "Optimize X-ray/ADP weight" option. All attempts end with a RMS(angle) of ~3.8 and a Clashscore of ~190.

Is there a known procedure to deal with this? I feel that my geometry weighing term is too loose but I don't know exactly how to tighten it other than decreasing the wxc_scale parameter. I have gone and manually corrected bad positions before but given the low resolution nature of the structure the clashes tend to be reestablished during refinement.

Thank you in advance for your help,

-- Mario Sanches Postdoctoral Researcher Samuel Lunenfeld Research Institute Mount Sinai Hospital 600 University Ave Toronto - Ontario Canada M5G 1X5 http://ca.linkedin.com/in/mariosanches

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

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

-- Mario Sanches Postdoctoral Researcher Samuel Lunenfeld Research Institute Mount Sinai Hospital 600 University Ave Toronto - Ontario Canada M5G 1X5 http://ca.linkedin.com/in/mariosanches

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

Sorry, I meant "optimeze x-ray/stereochemistry weight" (I ran both stereochemistry and ADP at one point to see what the effect would be). I always run real-space refinement on coot. I try to include a region around the residue(s) I am playing with to allow the structure to relax and to prevent local strain. I tend to select the "Ramachandran Restrains" and also select the appropriate secondary structure restraint when it is possible. I see regions with bad clashes solely because during refinement a side chain is being fit into a density peak that is to close to another residue atoms. I wonder if there is a way to tell the program to avoid clashes even if it is at the expense of real-space fitting? Mario Sanches On Thu, Apr 19, 2012 at 8:48 AM, Nathaniel Echols wrote:

...

I am refining a low resolution structure (2.9A) using phenix.refine. As I continue the refinement Rwork/Rfree are dropping but my geometry statistics in general are very bad. My clashscore is getting worst and

On Thu, Apr 19, 2012 at 7:37 AM, Mario Sanches wrote: the

...

RMS(angle) is really high. I have tried to play with the "wxc_scale" parameter and also ran phenix selecting the "Optimize X-ray/ADP weight" option. All attempts end with a RMS(angle) of ~3.8 and a Clashscore of ~190.

Did you mean to type "Optimize X-ray/Stereochemistry weight" instead? That is the option you want to use in these situations; it usually results in much better geometry, with two caveats:

1) If the input geometry is already poor, this may result in a poor initial guess for the weight, and the optimization will not sample the right range.

2) Severe clashes are hard to fix automatically - the structure needs to be repacked to fix them, which is beyond the power of our minimizer.

One tip: always run real-space refinement in Coot when rebuilding your model. Getting the input geometry as good as possible seems to help phenix.refine (or any other refinement program) converge on acceptable values.

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

-- Mario Sanches Postdoctoral Researcher Samuel Lunenfeld Research Institute Mount Sinai Hospital 600 University Ave Toronto - Ontario Canada M5G 1X5 http://ca.linkedin.com/in/mariosanches

Are you talking about the "auto fit rotamer" function? I am not using that. I tend to select rotamers from the list but I rarely pic the auto one (it tends to place my side chains inside the neighbouring main chain densities most of the time). Mario Sanches On Thu, Apr 19, 2012 at 9:06 AM, Nathaniel Echols wrote:

On Thu, Apr 19, 2012 at 8:00 AM, Mario Sanches wrote:

...

I see regions with bad clashes solely because during refinement a side chain is being fit into a density peak that is to close to another residue atoms. I wonder if there is a way to tell the program to avoid clashes even if it is at the expense of real-space fitting?

Are you using the automatic rotamer correction? This performs poorly at low resolution - 2.9A is about the point where it really breaks down.

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

-- Mario Sanches Postdoctoral Researcher Samuel Lunenfeld Research Institute Mount Sinai Hospital 600 University Ave Toronto - Ontario Canada M5G 1X5 http://ca.linkedin.com/in/mariosanches

Tim, the methods described in the paper you mention below are shown to work for some selected structures taken from PDB ( that is overall good structures). This is not Mario's case, unfortunately. It is not demonstrated that this methodology would work for poor models. All the best, Pavel On 4/19/12 10:05 PM, Tim Fenn wrote:

On Thu, Apr 19, 2012 at 5:37 AM, Mario Sanches wrote:

...

I am refining a low resolution structure (2.9A) using phenix.refine. As I continue the refinement Rwork/Rfree are dropping but my geometry statistics in general are very bad. My clashscore is getting worst and the RMS(angle) is really high. I have tried to play with the "wxc_scale" parameter and also ran phenix selecting the "Optimize X-ray/ADP weight" option. All attempts end with a RMS(angle) of ~3.8 and a Clashscore of ~190.

I'm guessing Pavel has already sorted this out - but an expert shouldn't be required to yield appropriate geometry/chemistry! This sort of problem has been discussed previously:

http://phenix-online.org/pipermail/phenixbb/2011-November/017906.html http://dx.doi.org/10.1107/S0907444911039060

although that high of a clashscore could arise from something else?

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