In recent years, deep learning-based methods have achieved significant breakthroughs in protein structure prediction. The open-source release of AlphaFold 2 (AF2) in 2021 enabled high-precision prediction of three-dimensional structures for both individual proteins and protein complexes, allowing researchers to rapidly obtain reliable structural information and greatly accelerating advancements in protein structure and function studies. The release of AlphaFold 3 (AF3) in 2024 took this further by achieving accurate predictions of three-dimensional structures for protein-nucleic acid and protein-small molecule complexes. With improved algorithms and a more efficient model, AF3 significantly enhanced prediction accuracy, especially demonstrating outstanding performance in antigen-antibody and protein-small molecule complexes. The success of AlphaFold has not only brought revolutionary progress to structural biology but also showcased immense application potential in fields such as drug development, protein design, and molecular function research, driving innovation in biomedical studies. This article will review the development history of AlphaFold and related protein structure prediction methods, summarize their key technologies and current applications, and, by considering their limitations, provide an outlook on future research directions and applications.