Accurate measurement of functional similarity between proteins is crucial for the prediction of protein-protein interactions and drug-target interactions. In this study, the functional similarity between proteins was computed using Gene Ontology (GO)....
Accurate measurement of functional similarity between proteins is crucial for the prediction of protein-protein interactions and drug-target interactions. In this study, the functional similarity between proteins was computed using Gene Ontology (GO). The similarity of terms was measured by word embedding methods, Word2vec and TF-IDF, which consider the term features and by a semantic similarity method that considers the relationship between terms. Thereafter, the measured similarity of terms was expanded to protein-protein similarity using annotations in GO. We proposed an integrated similarity that considers both features and relationships of the terms. Protein similarities measured by semantic similarity, Word2vec, TF-IDF, and integrated methods were applied to the prediction of protein-protein interactions and drug-target interactions. As a result, it was confirmed that the integrated similarity showed the best performance.