Terwilliger, Thomas C
terwilliger at lanl.gov
Fri Sep 6 16:29:44 PDT 2013
Oh I see. Yes, it is probably better to refine before building, either directly as you did or by setting refine_before_rebuild=True.
Autobuild really has just three ways of filling in those loops. One is running fit-loops (which it does in the middle of a cycle). Another is build_outside_model which it also does in the middle of a cycle. The third is just getting a better starting model so that chain extension fills in more residues. You can have autobuild just do build_outside model by turning off all the other building options but my guess is it is not worth it (if you really want to do this you can see how this is done by looking at a parameters file for a sub-process run of autobuild when only build_outside_model is run).
My guess is that the unbuilt part of the structure might not be really clear in the maps. If this is the case, then it is pretty hard to build that part of the model. Usually the best thing to do is get the best possible model (run autobuild, then phenix.refine, fix everything that is indicated by the Molprobity analysis and that you can see in the map), then use that in either refinement to get a 2Fo-DFc sigmaA-weighted map or phenix.autobuild maps_only=True to get a density-modified map. Then examine that map to see if there is anything that clearly should be built.
You note that "once I ran fit_loops, I would get an error if I tried to put the new pdb file through refinement.". If you could send me all the inputs and commands necessary to repeat that (or perhaps just the output PDB file) off-line then I can try and fix that problem.
All the best,
From: phenixbb-bounces at phenix-online.org [phenixbb-bounces at phenix-online.org] on behalf of Heather L Condurso [condurso at bc.edu]
Sent: Friday, September 06, 2013 5:14 PM
To: PHENIX user mailing list
Subject: Re: [phenixbb] Autobuild
First, I had refine input model before rebuilding turned off. I thought this meant whether it would refine before doing anything else which I didn't really think was necessary but maybe I was wrong. I found an autobuild.eff file and it looks like all that was different in my refinement was that I hadn't selected simulated annealing in autobuild and had in refine. I can run it again with these adjusted parameters to see if I get a better result.
As for the maps, the refine maps are definitely better than the autobuild maps but not by all that much. These missing regions are pretty long so fit_loops hasn't been too helpful. I also had the problem that once I ran fit_loops, I would get an error if I tried to put the new pdb file through refinement. Something about missing fields but it had only added a residue here and there across many monomers so I gave up.
I guess what would be helpful to me is how best to use my computing time to try to get autobuild to get me the best possible model it can. I can input my refined structure but what parameters will make autobuild focus on extending the model to build in the missing regions instead of continually trying to rebuild a model that is already at a CC=0.90 around the built atoms?
On Fri, Sep 6, 2013 at 6:32 PM, Terwilliger, Thomas C <terwilliger at lanl.gov<mailto:terwilliger at lanl.gov>> wrote:
Hmm... a little hard to tell what the problem is.
So first: autobuild should have refined this model as one of the first things it did. Did it get about the same R that you got when refining it directly with phenix.refine? If no, then have a look at the refinement .eff files (the autobuild one is in AutoBuild_run_1_/TEMP0 unless that was cleaned up) and see what is different.
Next....what happens if you just run phenix.refine to get a map, also take the map from this autobuild run, and look at them along with your model. Is one a lot better than the other?
If you take your best working map and run phenix.fit_loops with this model...how does that do?
All the best,
From: phenixbb-bounces at phenix-online.org<mailto:phenixbb-bounces at phenix-online.org> [phenixbb-bounces at phenix-online.org<mailto:phenixbb-bounces at phenix-online.org>] on behalf of Heather L Condurso [condurso at bc.edu<mailto:condurso at bc.edu>]
Sent: Friday, September 06, 2013 3:54 PM
To: PHENIX user mailing list
Subject: [phenixbb] Autobuild
I'm a bit confused about what is going on with my autobuild run. I solved a structure from anomalous data to 2.7A but many loops are missing. I ran phaser with the current model and native data that extends to 2.4A and got a solution and started autobuild from there. I used mostly the default settings and rebuild_in_place=False. After a pretty short time I realized the run was over and at the bottom of the log I see this:
Cycle 3 R/Rfree=0.28/0.35 Built=488 Placed=467
AutoBuild_build_cycle AutoBuild Run 162 Fri Sep 6 14:33:39 2013
Deciding if we should continue...
Recent R change per cycle: 0.0
Ending these build cycles as the recent change in R per cycle has been
0.0 and the required value of change is -0.005
and the R is somewhat acceptable at 0.276670644792
All done with build cycles in this region
AutoBuild_set_up_build AutoBuild Run 162 Fri Sep 6 14:33:40 2013
All omit/non-omit regions completed
finished AutoBuild Run 162 Fri Sep 6 14:33:40 2013
Finishing AutoBuild Run 162
Facts written to AutoBuild_run_162_/AutoBuild_Facts.dat
AutoBuild Run 162
Summary of model-building for run 162 Fri Sep 6 14:33:40 2013
Files are in the directory: /Users/heather/Desktop/M/AutoBuild_run_162_/
Starting mtz file: /Users/heather/Desktop/M/ImportRawData_run_4_/A-3_xx_PHX.mtz
Sequence file: /Users/heather/Desktop/M/m.pir
Best solution on cycle: 2 R/Rfree=0.26/0.33
Summary of output files for Solution 1 from rebuild cycle 2
--- Model (PDB file) ---
--- Model (CIF file) ---
--- Refinement log file ---
--- Model-building log file ---
--- Model-map correlation log file ---
--- 2FoFc and FoFc map coefficients from refinement 2FOFCWT PH2FOFCWT FOFCWT PHFOFCWT ---
--- Data for refinement FP SIGFP PHIM FOMM HLAM HLBM HLCM HLDM FreeR_flag ---
--- Density-modified map coefficients FWT PHWT ---
You might consider making one very good model now with:
My full sequence is 325 amino acids so only about 75% is build in. I am hoping autobuild can help find at least some of these missing residues. Why would the change in R be zero? What is meant by "You might consider making one very good model now with:"? I tried running autobuild with these new files and rebuild_in_place=true and get the same type of message. Is this the best that autobuild can do? I also ran the phaser solution through refinement and can significantly lower the Rs from 0.26/0.33 to 0.22/0.30. I appreciate any advice on how best to proceed.
I am using 1.8.2-1472 if that matters.
phenixbb mailing list
phenixbb at phenix-online.org<mailto:phenixbb at phenix-online.org>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the phenixbb