[phenixbb] autobuild mistake

Terwilliger, Thomas C terwilliger at lanl.gov
Wed Oct 23 11:57:01 PDT 2013

Hi Heather,

If you run a file like RUN_FILE_3.sh a second time, it will start a new clean sub-process...but then it will overwrite the file INFO_FILE_3 with the information from this second run.   AutoBuild uses the INFO_FILE_# files to figure out where the results are from the sub-processes.  In principle this should have worked (AutoBuild should have waited for your extra run to finish, should have combined the results from this extra run with the results from the 3 normal sub-processes, and ignored the 4th normal sub-process). I am guessing that perhaps you stopped your extra run (that would have been a pretty natural thing to do), and then AutoBuild was waiting for your stopped run to finish.

So yes, you can try recombining the models you got so far.  I'm puzzled by the R-values of 999.9 thought, as that usually means that refinement failed for some reason.

Try running a command like this:

phenix.combine_models mtz_in=AutoBuild_run_6_/TEMP0/resolve_work.mtz \
  pdb_in=AutoBuild_run_6_/TEMP0/AutoBuild_run_8_/overall_best.pdb \
  second_pdb_in=AutoBuild_run_6_/TEMP0/AutoBuild_run_9_/overall_best.pdb  \

which will take the best parts of these two models and put them together in one model (if you are lucky).  Then take the resulting combine_8_9.pdb and do it again with second_pdb_in as run_10...etc.

You can use whatever map file is the best one you have for this.

You can also try combining the models by running autobuild:

phenix.autobuild data=data.mtz model=model.pdb rebuild_in_place=False acceptable_r=0.2  consider_main_chain_list="AutoBuild_run_6_/TEMP0/AutoBuild_run_8_/overall_best.pdb  AutoBuild_run_6_/TEMP0/AutoBuild_run_9_/overall_best.pdb AutoBuild_run_6_/TEMP0/AutoBuild_run_10_/overall_best.pdb AutoBuild_run_6_/TEMP0/AutoBuild_run_11_/overall_best.pdb"

which should use all those models as information for model-building.

The combine models keyword in autobuild is better suited for a rebuild_in_place run so that is not so appropriate.
I'm not too optimistic however given the R values...

All the best,
Tom T

From: phenixbb-bounces at phenix-online.org [phenixbb-bounces at phenix-online.org] on behalf of Heather L Condurso [condurso at bc.edu]
Sent: Wednesday, October 23, 2013 8:11 AM
To: PHENIX user mailing list
Subject: [phenixbb] autobuild mistake

So I was running autobuild on the command line with:

phenix.autobuild data=data.mtz model=model.pdb rebuild_in_place=False acceptable_r=0.2

I have a model from another data set that is missing some loops and new data sets which I would like to use autobuild to see if the model can be extended.

I am using 1.8.4-1496 on a mac with osx 10.6.8.  I was an idiot and when looking in the directory to check out one of the RUN_FILE_#.log files, I accidentally clicked on the RUN_FILE executable which started another AutoBuild subprocess.  Since then all the other subprocesses had finished, but overall run stalled at:

Splitting work into 4 jobs and running with 1 processors using sh
background=False in ./AutoBuild_run_6_/TEMP0
Final job will be run with sh  with background=True
Starting job 1...Log will be: ./AutoBuild_run_6_/TEMP0/RUN_FILE_8.log
Starting job 2...Log will be: ./AutoBuild_run_6_/TEMP0/RUN_FILE_9.log
Starting job 3...Log will be: ./AutoBuild_run_6_/TEMP0/RUN_FILE_10.log
Starting job 4...Log will be:  ./AutoBuild_run_6_/TEMP0/RUN_FILE_11.log

And the combine and extend didn't happen.  I saw a post about combining models from autobuild but it was for rebuild_in_place=True and the post mentioned that the models needed to have identical sequences for it to work.  Is there a way to save my mistake without starting over?

The run I had restarted had yielded Rs=999.9 so If there is a way to maybe combine just 3 of the 4 that may be helpful as well.

Heather Condurso
condurso at bc.edu<mailto:condurso at bc.edu>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20131023/b06a9860/attachment.htm>

More information about the phenixbb mailing list