Artificial Intelligence Powers Protein Functional Annotation

Abstract

Protein functional annotation is essential for understanding biological processes, disease mechanisms, and enzyme activities, yet experimental validation remains costly and low-throughput. With the rapid development of Artificial Intelligence (AI), a wide range of computational approaches have been proposed to infer protein functions. This review systematically examines methods for annotating Gene Ontology (GO) terms and Enzyme Commission (EC) numbers. These are two complementary systems that capture different aspects of protein functions. Based on these two systems, we first synthesize existing approaches into six general modeling paradigms with a clear, structured framework. Then, we introduce GO and EC in a parallel manner, consisting of representative methods, commonly used evaluation metrics, prediction scenarios, and task-specific challenges. Finally, we outline emerging opportunities and future directions aimed at achieving more accurate, context-dependent, and high-resolution protein functional annotation.