You are here

Problem about energy score

8 posts / 0 new
Last post
Problem about energy score

FlexPepDock Refinement;Score.
Hi all!
I am trying to dock a peptide to a receptor using FlexPepDock Refinement and generated 600 models,my top five models lie within sub-angstrom RMSDbb of the native structure.The lowest energy is -557.578.
Here is my flags:
-s pm6o.pdb

-native native.pdb





-nstruct 600

Performing the command:
score.linuxgccrelease @flags > score.log
I got the total score of the native structure(-316.451).I'm confused that the total score of the native is greater than the models(-316.451>-557.578).Does it means there is something wrong with the scoring function(SCOREFUNCTION:standard SCOREFUNCTION PATCH:score12) or the default scoring function for FlexPepDock search is unfitted. My files have been attached.
Any replies is greatly appreciated.

top five models score.jpg68.31 KB
native score.txt619 bytes
Post Situation: 
Tue, 2013-11-05 02:09


This is a common observation: the crystal structure (native structure) is usually scored with less favorable, higher energy than any structure optimized using the Rosetta forcefield.
You should compare therefore your models to a near native structure obtained by local optimization, which would change the structure minimally, but reduce the energy significantly, by e.g. relieving clashes.

For FlexPepDock applications, you may use the flag -flexPepDockingMinimizeOnly to do so. More details can be found in the manual at


Tue, 2013-11-05 02:33
Ora Schueler-Furman

Thank you very much for your response Ora.
You're right.I use the flag -flexPepDockingMinimizeOnly to perform a short minimization of the native structure and the energy of native structure reduces obviously(total_score:-494.152),but is still higher than the models.I wonder if i can modify the default scoring function and how to modify it.

Tue, 2013-11-05 06:07

Before changing the scorefunction, I'd check to see what's giving you the difference. If you have your models output as PDBs, there should be a table at the end of the PDB which shows the energies for each residue broken down by scoreterm. You can try comparing the two tables to see if it's a particular residue/scoreterm which is giving you the difference. Once you know that, you can examine the native and the docked structures to see if you can rationalize the difference. This hopefully will give you insight as to what you should be doing differently.

Tue, 2013-11-05 07:03

Just to add to the discussion: Which version of Rosetta are you using? The default energy function and settings changed from 3.5 to the more recent weekly releases (talaris2013). If you are using a weekly, to get the same options for standard/score12 you will need to pass the option -restore_pre_talaris_2013_behavior

You may also want to look into relaxing your native structure first using the relax application with the option -constrain_relax_to_start_coords. This will use the more rigorous minimization of the relax algorithm, while not messing with your structure too much.

Tue, 2013-11-05 07:44

could you please upload the the native or the optimized native and one of the decoy structures, so I can look into it?


Tue, 2013-11-05 09:27
Ora Schueler-Furman

The version of Rosetta i use is 3.4.Here is my PDB files and score files.

Thu, 2013-11-07 00:59

In order to compare structures that differ only at the interface, you need to use the same background (or look at different score terms, such as peptide score or interface score only). In your case it looks like you did not prepack the native when you did the optimization run. Therefore the receptor side chains might differ. I can see residues in the native structure that have unfavorable energy values resolved in the decoy (e.g. high repulsive value for His 263 that is flipped in the prepacking step) and are not at the interface.


Thu, 2013-11-07 03:25
Ora Schueler-Furman