Hi Foggy,
Here is another thing you can try:
Ignore the chain B completely for now, just you would a ligand. Run autobuild with rebuild_in_place=False, and then afterwards with rebuild_in_place=True to get the best model you can. Even without 10% of the model I'd expect you might get a bit better model than you have now that way.
Then focus on chain B. If you cannot see something fairly clearly in your map then autobuild won't be able to either. So if there is clear density for chain B, autobuild should build it using the methods Kostya and I suggested before. You can also use phenix.cut_out_density to just cut out the density for chain B and then you can build into that density with phenix.find_helices_strands or phenix.build_one_model, then put it back in the unit cell.
All the best,
Tom T
________________________________
From: [email protected] [[email protected]] on behalf of Konstantin Kogan [[email protected]]
Sent: Sunday, September 14, 2014 12:18 AM
To: Zhang Foggy
Cc: [email protected]
Subject: Re: [phenixbb] [ccp4bb] Autobuild in specific area
Hi Foggy,
I had similar situation recently. I would suggest to build manually or in coot (place a helix) at least one helix that you see clearly. Then check that protein a and protein b (helix) are different chains in model pdb. Then provide the sequences of both chains. In this situation autobuild will certainly rebuild this helix with real sequence and will try to build more (don't forget to select 'build outside model' option). I would also increase the number of rebuilding cycles to something 20-30. In my case it took around 7 cycles until it started to put any new amino acid.
Kostya
On Sep 14, 2014, at 7:51 AM, Zhang Foggy wrote:
Hi, Tom,
Thanks for your great suggestions.
Here is my awkward situation: protein-A has 900 amino acids, while protein-B only 70. When I tired your first suggestion, I found that the program only built up the protein-A part. No extra model was built up at all.
Additionally, there is no published model for protein-B. Sequence alignment and homolog modeling indicate that there are about 40 aa forming two alpha-helix with the rest unknown. I can only see clear density on one alpha-helix at the moment, and the R/Rfree values are 0.37/0.43 (2.5 A, P1). Regarding to your second suggestion, I do not find this option though GUI yet, will try tomorrow. Do you have any other suggestions on my situation? Thank you in advance.
Best,
Foggy
2014-09-13 21:03 GMT-05:00 Terwilliger, Thomas Charles