Confidence score and constraints

[tkramer-motion]

Is the confidence score influenced by if the atom positions meet the given constraints? If so is there a way to generate a "global" confidence using the atom positions generated with the constraints?

[wukevin]

Is the confidence score influenced by if the atom positions meet the given constraints?

The confidence scores are designed to be a holistic estimate, so they are indeed influenced by the restraints.

If so is there a way to generate a "global" confidence using the atom positions generated with the constraints?

While there is no single best "global confidence" with regard to a constraint, there are several ways to go about this depending on what "higher order" measurement you are trying to understand. At its simplest, one could simply check whether the predicted structure adheres to the given restraints, and could then get a boolean score that way. You might also consider things like the PAE restricted to the restraint you have specified (e.g., the error of one partner in the restraint aligned to the frame of the other partner, perhaps including a few neighboring residues to amortize out some noise, etc.).

Let us know if you have a more specific example/use case and we can brainstorm what might make more sense.

[arogozhnikov]

confidence score consists of several terms, but neither of those explicitly account for satisfaction of restraints