Dear Sebastian and PhenixBB users,

I would like to post short summary that could help other users in similar situation. When using not very similar reference model (e.g. sequence similarity is below 85%) it can be useful to define particular selections that are similar using Settings (menu) -> Reference model -> Reference group... in the GUI.

If this is not enough, couple parameters are available in the latest nightly build (with defaults):

refinement.reference_model.search_options.min_contig_length=10
refinement.reference_model.search_options.min_percent=85

If you are using command line, you can put them in parameter file or in command line. In GUI you can find them in refinement settings -> All parameters... -> search parameters, using 'search_options' there.
min_contig_length value can be set to 0,
min_percent, which is sequence similarity in percents also could be set to lower values.

I would like to thank Sebastian Falk for bringing this issue to our attention and cooperation in solving this problem.

Please note, that for majority of cases default values should be good enough. The names of these parameters may be changed in the feature.

Best regards,
Oleg Sobolev.


On Tue, Dec 22, 2015 at 9:09 AM, Oleg Sobolev <osobolev@lbl.gov> wrote:
Dear Sebastian,

I hope you are using at least latest official release of Phenix.

One of the possible ways to proceed is to select only parts of your reference structure that have more sequence identity and supply them along with corresponding parts of your model via reference_model.reference_group scope. I'm sure that there is a way to do so in Phenix GUI as well.

I'm happy to provide more specific advice, but I will need your files to do so. You can send them directly to me (off-list). They will be kept in confidence.

Best regards,
Oleg Sobolev.

On Tue, Dec 22, 2015 at 2:28 AM, Falk, Sebastian <sfalk@biochem.mpg.de> wrote:
Hi!

I am currently building a 3.3 A structure of a large protein. I do have a reference model at much better resolution that has 58% sequence identity and 77% similarity.
The structures are quite similar, there are only a few regions with insertions.
However, phenix.refine does ‘accept’ the reference model and thus creates no torsion restraints.

I also tried to run ProSmart in CCP4 but there the cut-off for the reference model is 75% sequence identity. I could not find any threshold for phenix.refine though.

I am happy for any suggestion or tip how to include the reference model even though the sequence identity is not as high.

Thank a lot.

Sebastian






Sebastian Falk
Max-Planck-Institute for Biochemistry
Structural Cell Biology
Am Klopferspitz 18
D-82152 Martinsried
Germany

sfalk@biochem.mpg.de


_______________________________________________
phenixbb mailing list
phenixbb@phenix-online.org
http://phenix-online.org/mailman/listinfo/phenixbb
Unsubscribe: phenixbb-leave@phenix-online.org