In the original paper the authors propose a new evaluation that respects inter-novel-protein interactions, and also a new method, that significantly outperforms previous PPI methods, especially under this new evaluation. We first confirmed that the new evaluation protocol is indeed much better in assessing the models generalization abilities, which was the main problem of the field that the authors were trying to solve. Secondly, we tried to reproduce the results of the proposed model in comparison with previous state-of-the-art, PIPR. Our results show that it really did perform better, but in some experiments the uncertainty was too big, so some claims were only partially confirmed.