AlphaFold2 and its applications in the fields of biology and medicine

Abstract

AlphaFold2 (AF2) is an artificial intelligence (AI) system developed by DeepMind that can predict three-dimensional (3D) structures of proteins from amino acid sequences with atomic-level accuracy. Protein structure prediction is one of the most challenging problems in computational biology and chemistry, and has puzzled scientists for 50 years. The advent of AF2 presents an unprecedented progress in protein structure prediction and has attracted much attention. Subsequent release of structures of more than 200 million proteins predicted by AF2 further aroused great enthusiasm in the science community, especially in the fields of biology and medicine. AF2 is thought to have a significant impact on structural biology and research areas that need protein structure information, such as drug discovery, protein design, prediction of protein function, et al. Though the time is not long since AF2 was developed, there are already quite a few application studies of AF2 in the fields of biology and medicine, with many of them having preliminarily proved the potential of AF2. To better understand AF2 and promote its applications, we will in this article summarize the principle and system architecture of AF2 as well as the recipe of its success, and particularly focus on reviewing its applications in the fields of biology and medicine. Limitations of current AF2 prediction will also be discussed.

Fig. 1

Performances of protein structure prediction indicated as backbone agreement with that of structures determined by experiments for the best models in CASPs. a The trend of performance (denotated by GDT_TS values) with regard to the backbone accuracy for best models obtained in each CASP. b A comparison of the GDT_TS values with or without AF2 prediction included in CASP14. Prediction accuracies for proteins with different target difficulty (“easy”, “medium” and “difficult”) are presented in indicated colors (blue, gold and green, respectively)

Fig. 2

Schematic work principle and architecture of AF2. a The overall architecture of AF2. The pipeline of AF2 contains three modules. The first one is the input module, which takes an amino acid sequence as input, and generates the MSA representation and the pair representation. The second one is the Evoformer module, which takes the MSA representation and the pair representation from the first module and passes them through the deep learning module, Evoformer. The third one is the structure module, which achieves the transition from abstract representation of protein structure to 3D atom coordinates of target protein. b Components of a block in Evoformer. Evoformer contains 48 blocks with weights not shared. The MSA representation and the pair representation are renewed through each block. c Components of a block in the structure module. Structure module contains 8 blocks with shared weights. Single representation and backbone frames are updated through each block of the structure module

Fig. 3

Application areas of AF2 in the fields of biology and medicine. AF2 can be applied in many areas of biology and medicine, including structural biology, drug discovery, protein design, protein-protein interaction, target prediction, protein function prediction, biological mechanism of action, and others (such as protein evolution, rare disease treatment studies, effects of mutation on treatment, vaccine design and so on)