You are here

Antibody scoring

2 posts / 0 new
Last post
Antibody scoring


I am wondering what the scores that are provided in the score file of an antibody job mean. Does a more negative score mean that a model is expected to be closer to the native structure? 


Post Situation: 
Wed, 2022-06-22 11:14

Yes, generally speaking more negative Rosetta scores (energies) are considered to be better.

There's some caveats there, in that the Rosetta score function is not an exact match to the natural energy landscape. So it's possible that a low scoring structure is at a false minimum. In practice, these spurious minima tend to be rather narrow (versus native minima being broad), so you can often tell if a low-energy structure is a false minimum by looking at the clustering results (are there a number of slightly different structures all with good scores around the low-energy structure), or the related concept of looking at a score-vs-rmsd "funnel plot" (using a low energy structure as an rmsd reference): if the reference structure is native like, there should be a "funnel" to it, with the bottom most (most negative energy) structures at a given rmsd generally getting higher and higher energy as the rmsd increases.

If you're looking for more information about what all the subscores being reported are, see

Wed, 2022-06-22 12:00