RMS determination using Rosetta 3.3

[Anonymous]

To give some context, I have a set of decoys that I’ve run cluster analysis on and now I’m trying to look at analyzing the clusters to see what similarities there are within the cluster. The long term goal is to analyze the 0th cluster, which contains the most structures, which is hypothesized to be the most likely place for where the true structure will be. That being all fine and dandy, I’ve isolated the poses that I need and was trying to determine the rms of each pose against the middle pose of the cluster. This was supposed to be a 15 minute exercise in double checking assumptions, but it became more complicated.

I used the ‘score.linux.gccrelease’ script with the following flags:

-in:file:l listofcluster.txt
-in:file:native center.pdb
-out:file:scorefile scorefile.txt

I’ve removed a lot of the directory pointing and renamed from what my file names are, but that’s the gist of it. At that point everything seemed to be working fine. In the output I had a slew of information, including columns for rms and allatom_rms. In hopes of lining up the poses with the corresponding values, I discovered that I had more values than I had poses. I then tried getting a smaller cluster and retrying the whole process. With a 24 pose cluster I received 35 pairs of values(rms, fullatom_rms), two of which had rms and fullatom_rms equal to zero.

If there is anything else that would be helpful in solving the problem, let me know.

[Anonymous]

Scorelines are typically appended to the scorefile, so if you run the application multiple times, you’ll get an accumulation of scoring lines from each of the multiple runs. My guess is that your scorefile with 35 lines is actually getting lines from two runs, one with 24 poses and one with 11 poses (so you get the zero rmsd from each of the centers). Another possibility is that your application is getting interrupted halfway through and restarted, resulting in an accumulation of lines from both runs. (A third possibility, though unlikely, as you don’t have an even multiple of the input structure number, is that you’re somehow setting -nstruct and getting multiple rescores of each input structure.)

Try manually re-running the scoring in a clean directory, and check the description column on each line of the resultant scorefile. There should be one (and only one) line for each of the input files.

[Anonymous]

Thanks, you are absolutely correct. I’m not sure how we managed it, but we did have two runs with different numbers of poses. Of all the silly errors…

Thanks again. Your help is much appreciated.

Mark

[Author]

Posts