PPaxe was developed at the Computational Genomics Lab, UB. You can find the code of the website here, and for PPaxe here.

In order to retrieve the interactions, PPaxe uses the Stanford CoreNLP for entity recognition and a Random Forest Classifier trained to recognize interacting proteins at the Computational Genomics Lab. The performance of this predictor (evaluated using the 10-fold cross validation method) can be seen below:

ACCURACY	84.48
PRECISION	76.50
RECALL	58.68
F1	66.41
AUC	0.92

---