That should explain most of the standard scoring terms. If there's a particular entry in the score file that's not covered, feel free to ask about it specifically.
You should use the I_sc term to pick models. total_score is dependent on the monomer models as well, so it does not discriminate as well. I_sc is the proxy for binding energy in RosettaDock. If you're running the low resolution mode as well with motif dock score (highly recommended; much better than the default centroid score function), the motif_dock_score column is also a very good discriminant.
You need to set the native structure to get CAPRI-related metrics because they are comparison metrics. The method doesn't assign a native on its own. Without a native structure specified at the start, it will give either NaN or nonsense values for the Irms, Fnat, CAPRI-type columns.
The general reference for Rosetta score terms is Alford et al. https://pubs.acs.org/doi/full/10.1021/acs.jctc.7b00125
That should explain most of the standard scoring terms. If there's a particular entry in the score file that's not covered, feel free to ask about it specifically.
Thank you very much for your explanation.
I have looked through the paper.
What I am wondering is if those terms with prefixes like fa_XXX or hbond_XXX measure only one aspect of the energy.
If I want to pick up better docking models, should I choose the function that combines all of the terms (e.g., total score)?
Another thing I want to confirm is that the score terms related to CAPRI require a reference model (native model).
I wonder whether the prediction with the lowest energy will be used as a reference.
If so, does that mean those CAPRI-related score terms are dependent on the selection of the reference model?
Thank you in advance.
You should use the I_sc term to pick models. total_score is dependent on the monomer models as well, so it does not discriminate as well. I_sc is the proxy for binding energy in RosettaDock. If you're running the low resolution mode as well with motif dock score (highly recommended; much better than the default centroid score function), the motif_dock_score column is also a very good discriminant.
You need to set the native structure to get CAPRI-related metrics because they are comparison metrics. The method doesn't assign a native on its own. Without a native structure specified at the start, it will give either NaN or nonsense values for the Irms, Fnat, CAPRI-type columns.