Machine Learning for Protein Function Prediction

Knowledge of protein functions is crucial to understanding and investigating cellular functions across all organisms. Accurate annotations of protein functions are also useful for the elucidation of mechanisms of various diseases and can be used to guide target-based drug design efforts. Although biological experiments are the most precise way for functional annotation of proteins, they are often time-consuming, laborious, and expensive. Therefore, there is an urgent need to develop efficient and accurate computational approaches for protein function prediction. This chapter comprehensively reviews and categorizes prominent computational predictors of protein functions, which are defined by the Gene Ontology (GO) terms, including template detection-based methods, statistical machine learning-based methods, deep learning-based methods, and composition methods. Applications of those protein function prediction methods are also discussed.