Two years after its initial release, AlphaFold2 has become a widely adopted tool for protein structure prediction. This review discusses recent developments and applications of AlphaFold2, focusing on its use in structural biology. AlphaFold2 has significantly accelerated experimental structure determination by enabling faster and more accurate predictions, which are used in techniques like molecular replacement and cryo-EM. It has also enabled the discovery of new protein-protein interactions and the design of novel proteins. AlphaFold2's predictions are continuously validated against experimental structures, helping to refine its accuracy and identify its limitations. AlphaFold2 has had a major impact on cryo-EM, allowing for the integration of predicted structures with experimental data to improve resolution and accuracy. It has also been used to identify unknown densities in cryo-EM maps through structural search. In protein design, AlphaFold2 is used as a component in design pipelines to generate initial models, which are then refined. It has also been used to predict protein-protein interactions, including those involving protein-peptide complexes. AlphaFold2 has enabled the construction of large prediction databases, such as the AlphaFold Database, which now contains over 200 million UniProt sequences. These databases have been used to develop efficient algorithms for structural search and to improve the understanding of protein folds. AlphaFold2's confidence metrics are also used to evaluate the accuracy of its predictions, with higher confidence models being more reliable. Recent studies have shown that AlphaFold2 can predict protein structures with high accuracy, even in cases where experimental data is limited. However, its predictions are not always accurate, and lower confidence models may require experimental validation. The ongoing development of AlphaFold2 continues to expand its capabilities, including the prediction of protein-DNA and protein-RNA complexes, and the elucidation of the effects of point mutations. The future of AlphaFold2 includes improving predictions for larger complexes, enhancing the accuracy of predictions for antigen-antibody interactions, and refining domain positioning for membrane proteins. The integration of AlphaFold2 into structural biology workflows has transformed the field, enabling new research directions and applications. The widespread adoption of AlphaFold2 highlights its importance in structural biology and its potential to further advance the field of protein structure prediction.Two years after its initial release, AlphaFold2 has become a widely adopted tool for protein structure prediction. This review discusses recent developments and applications of AlphaFold2, focusing on its use in structural biology. AlphaFold2 has significantly accelerated experimental structure determination by enabling faster and more accurate predictions, which are used in techniques like molecular replacement and cryo-EM. It has also enabled the discovery of new protein-protein interactions and the design of novel proteins. AlphaFold2's predictions are continuously validated against experimental structures, helping to refine its accuracy and identify its limitations. AlphaFold2 has had a major impact on cryo-EM, allowing for the integration of predicted structures with experimental data to improve resolution and accuracy. It has also been used to identify unknown densities in cryo-EM maps through structural search. In protein design, AlphaFold2 is used as a component in design pipelines to generate initial models, which are then refined. It has also been used to predict protein-protein interactions, including those involving protein-peptide complexes. AlphaFold2 has enabled the construction of large prediction databases, such as the AlphaFold Database, which now contains over 200 million UniProt sequences. These databases have been used to develop efficient algorithms for structural search and to improve the understanding of protein folds. AlphaFold2's confidence metrics are also used to evaluate the accuracy of its predictions, with higher confidence models being more reliable. Recent studies have shown that AlphaFold2 can predict protein structures with high accuracy, even in cases where experimental data is limited. However, its predictions are not always accurate, and lower confidence models may require experimental validation. The ongoing development of AlphaFold2 continues to expand its capabilities, including the prediction of protein-DNA and protein-RNA complexes, and the elucidation of the effects of point mutations. The future of AlphaFold2 includes improving predictions for larger complexes, enhancing the accuracy of predictions for antigen-antibody interactions, and refining domain positioning for membrane proteins. The integration of AlphaFold2 into structural biology workflows has transformed the field, enabling new research directions and applications. The widespread adoption of AlphaFold2 highlights its importance in structural biology and its potential to further advance the field of protein structure prediction.