- This topic has 8 replies, 4 voices, and was last updated 12 years, 2 months ago by Anonymous.
March 20, 2011 at 12:23 pm #835Anonymous
I tried Rosetta (mr_rosetta) to solve a difficult molecular replacement problem on my dataset, by following the demo inside released rosetta 3.2 software. I got questions about the README file:
In the 5th step, I generate 2000 models building all gaps up to length 8. According to the README, ” The top 10% are selected by Rosetta energy; each of these structures are rescored with phaser; the best is relaxed again using the updated density”. This means I need to use phaser to rescore 200 structures,which I can just write a shell script to do and is not a problem for me. The question is afterwards how many percentage of 200 structures I can define to “best”. Maybe only top 10?
I appreciate if somebody, who have tried or know how to go through the whole process, can give me some valuable suggestions. Thanks.
March 21, 2011 at 1:36 pm #5206Anonymous
One more question is how to get the updated density for relaxing?
March 21, 2011 at 2:42 pm #5211Anonymous
I have emailed someone knowledgeable to try to bring this to their attention.
March 21, 2011 at 6:34 pm #5219Anonymous
I’ve already got valuable suggestions from Frank, thanks frank and smlewis.
March 21, 2011 at 6:54 pm #5221Anonymous
Reposting for reference:
For previous blind cases, I rescored the top 10 percent (200 models) with phaser. At that point, if the correct MR solution is among the 200, there should be a very clear score separation … in other words, the top-scoring models would all look very similar (and be in the same orientation). If they do not, then the correct MR solution was probably not in the input set (alternatively, there may be errors in the sequence alignment). In cases where the top models are in agreement, I would generally take 5 or fewer models through the additional relax (often just 1).
To build the density map phased the model, I used the phenix.maps program (http://www.phenix-online.org/documentation/maps.htm). You might also want to take a look at the phenix interface to mr_rosetta, http://www.phenix-online.org/documentation/mr_rosetta.htm.
May 25, 2011 at 8:56 am #5639Anonymous
June 8, 2011 at 7:38 pm #5739Anonymous
I replied to the user directly, but I am reposting the solution below:
The LLG dropping indicates something is going wrong. I think the problem is the sequence identity you give to phaser. Even after refinement in Rosetta, you should keep this value at its original sequence identity (25% or whatever). Using this same value, you should see an increase in LLG.
September 13, 2011 at 11:12 am #6034Anonymous
I posted this on the PyRosetta forum, but perhaps some people on this thread might be able to help me.
Is there a way to do these refinements within PyRosetta (refining a PDB structure based on low-resolution electron density from XRay crystallography or cryoEM)? I was hoping that there would be a possibility to include this in PyRosetta, so that users can interface the I/O with Numpy and Scipy. Do you know a way to implement this in PyRosetta? Is anything planned in this direction?
Any advice, whether positive or negative will be appreciated.
May 25, 2011 at 8:56 am #5640Anonymous
During last several weeks, I go back to use this method, trying to get a nicer solution. In my case, the space group is C121, with 2 or 3 copies in the ASU. I tried 2 copies first and from the 2nd step after I place the template into the intensities, I got a low positive LLG score (20) and low TFZ (4.7). Then I go to 4th step to do CM into density. After this step I did rescoring on the models coming out from the 4th step,using the script from the 2nd step with a higher sequence identity (not 100% since there are still missing part in the model). However, this time the LLG goes very negative (-600). Does this mean the “CM into density” goes to somewhere very dark? I appreciate if you have time and give me some suggestions on my case. Thanks.
- You must be logged in to reply to this topic.