Identifying interactions between proteins is important to understand\nunderlying biological processes. Extracting a protein-protein interaction (PPI)\nfrom the raw text is often very difficult. Previous supervised learning methods\nhave used handcrafted features on human-annotated data sets. In this paper, we\npropose a novel tree recurrent neural network with structured attention\narchitecture for doing PPI. Our architecture achieves state of the art results\n(precision, recall, and F1-score) on the AIMed and BioInfer benchmark data\nsets. Moreover, our models achieve a significant improvement over previous best\nmodels without any explicit feature extraction. Our experimental results show\nthat traditional recurrent networks have inferior performance compared to tree\nrecurrent networks for the supervised PPI problem.\n