search

FreshAirTonight / af2complex

Predicting direct protein-protein interactions with AlphaFold deep learning neural network models.
146 stars 19 forks source link

Calculating the predicted interface TM-score and the interface score. #2

Closed IlyesAbdelhamid closed 2 years ago

IlyesAbdelhamid commented 2 years ago

Hello,

I would like to know whether it is possible to calculate both the predicted interface TM-score and the interface score from the features.pkl only ? I am looking for a way to extract just what it is required for the calculation of these two metrics without running the entire AF2 pipeline. Thank you in advance for your help.

Sincerely, Ilyes

FreshAirTonight commented 2 years ago

Ilyes,

The interface score or interface TM-score is calculated based on a predicted structure model, specifically, the result (output as a pickle) from AF2 deep learning inference. It is not possible to calculate these two scores directly on features.pkl w/o through the model inference step. However, with the features.pkl, you may run AF2Complex to predict structure models and get the scores you need, w/o going through the feature derivation step (skipping all sequence library searches). Af2Complex works with features of either individuals monomers or of a multimer feature set. Please see the examples for both cases.

IlyesAbdelhamid commented 2 years ago

Thank you very much for your prompt reply. I managed to include the metrics in AF v2.1.2 and compute them. Thousand apologies for the delay!