You are here

calculate low-resolution total docking score for a input pdb

4 posts / 0 new
Last post
calculate low-resolution total docking score for a input pdb
#1

I have generated many low resolution decoys from a low-resolution docking run and wish to compare the total scores with whatever the comparable total score would be for the starting structure (which is input as full atom structure).  So, the startiing structure would need to be converted to centroid form.  Is there any way get the total score for just one input structure, in centroid form without generating decoys.  I know that when I use Rosettadock clusttering on the set of decoys, (in silent mode), the total scores are output for each pdb structure, which are identical to the original scores from the low-res docking, but those pdb structures are in centroid mode.  So, perhaps a trick is to convert my starting structure to a centroid form (How to do that ?) and then just pick some set of the decoys along with the centroid starting structure, and run quick clustering, which presumably would give some kind of total score for the starting structure.

Thoughts?

Thanks in Advance.

J Snyder

Category: 
Post Situation: 
Fri, 2015-12-11 23:08
jasnyderjr

If you use the -in:file:centroid  option, PDBs will be read in as centroid mode Poses. If you just want to get centroid mode scores, you can pass this option to the score_jd2 application, along with the "-score:weights" option specifying which centroid mode scorefunction you want to use to rescore, and it should output the centroid mode scores.

If you actually want to convert the structures to centroid mode, you can use score_jd2 with the "-out:pdb" or the "-out:file:silent" options, along with the -in:file:centroid option, to actually output the (rescored) structures in centroid mode form. (But keep in mind that Rosetta will normally treat centroid PDBs as fullatom PDBs missing their sidechains, unless you pass the -in:file:centroid option when you read them back in.

Tue, 2015-12-15 10:04
rmoretti

Very much appreciate the response:

I just picked a output decoy pdb from a low-resolution only run, and followed the suggestion  ("score_jd2")  However, I get totally different sore values compared to the score from the original output file:

 

Using score_jd2 with:   -in:file:centroid   -score:weights cen_std   -s prepack__1ahw_0001_03847.pdb   -out:pdb

and get for output: 

SCORE: total_score       cbeta         env linear_chainbreak overlap_chainbreak               pair                vdw description
SCORE:     -38.259     218.208    -139.559             0.000              0.000           -118.028              1.120 prepack__1ahw_0001_03847_0001

 

From output score file of original  low-resolution run:

 

SCORE: total_score         rms     cen_rms interchain_contact interchain_env interchain_pair interchain_vdw        st_rmsd description
SCORE:     -79.655      67.025      67.025            -20.000        -59.146          -0.509          0.000         62.120 prepack__1ahw_0001_03847

 

Obviously, different terms are involved. I'm aware that there are several score functions in use for different stages of protocols,

The pertinent options used with the low-resolution run (docking_protocol.default.linuxgccrelease) were:

-ex1
-ex2aro
-native input/1ahw.pdb
-docking:partners AB_C
-nstruct 30000
-docking:low_res_protocol_only
-randomize1
-mute core.util.prof ## dont show timing info
-out:file:scorefile 1ahw.fasc
-mute core.io.database
-use_input_sc
 

 I also tried -score:weights score3 and got much differenent results from above two outputs.


So, I noticed some related forum addresses to this question, but I feel need to ask.  If I wish to compare scores from the low-resolution decoys with that for a native structure, I would need to have same terms used by the score_jd2--have everything on same level, assuming this is possible.

 

Thanks Again,

J Snyder

Wed, 2015-12-16 00:15
jasnyderjr

You're right - if you want to match the scoring, you need to make sure you're using the correct weights. From what I can tell, it looks like the scores for your low resolution runs are using the "interchain_cen" weight set. Pass that to -score:weights and you *should* get close to the same energies as before.

Wed, 2015-12-16 11:22
rmoretti