DeepGO is a novel method for predicting protein functions using protein sequences and protein-protein interaction (PPI) networks. It uses deep neural networks to learn sequence features and PPI embeddings learned with knowledge graph representation learning method and hierarchically classifies proteins with GO classes.