search

nrbennet / dl_binder_design

MIT License
203 stars 49 forks source link

af2_complex_rmsd score? #58

Closed caleblareau closed 9 months ago

caleblareau commented 9 months ago

Apologies if I missed this somewhere, but is there an easy way to get the RMSD following the execution of the dl_binder_design/af2_initial_guess/predict.py script? Or does that require running another piece of the workflow? I see the various plddt and pae outputs in the .silent file but not the af2_complex_rmsd metric discussed in the methods. Thanks!

nrbennet commented 9 months ago

I have now added this. This gets saved alongside the other scores like pae_interaction. There is binder_aligned_rmsd which is the RMSD of the predicted binder aligned to the initial design model of the binder. And there is also target_aligned_rmsd which is the RMSD of the predicted binder when the two complexes are aligned on the target structure, this is a more stringent alignment metric and is the same as what is calculated in af2_complex_rmsd

caleblareau commented 9 months ago

awesome-- thanks for adding the feature and clarifying the metric.

KoningCheems commented 6 months ago

@nrbennet Just to clarify

  1. The target_aligned_rmsd You first align the initial and final targets, and then calculate the RMSD of the binder only? So it is more stringent because if they are actually far away from each other (i.e not binding) then the score will be large?
  2. The binder_aligned_rmsd You first align binders and then calculate the rmsd of the binders only?