When it comes to AlphaGo, most people have more or less heard that it was praised as a "world feat" because it defeated the human Go master, marking an important milestone step towards universal AI. Now, the AlphaFold2 model in cutting-edge scientific research is another symbol of the application of artificial intelligence (AI). It is favored by people from all walks of life because it shows the great potential of AI for Science. AlphaFold2 helps to predict protein structure. AI for Science new high protein is an important component of human cell tissue. Effective analysis and prediction of protein three-dimensional structure can provide important basis for future research and development of biology, medicine, pharmacy, agriculture, animal husbandry and other industries, especially for drug research and development directly related to human health. However, traditional methods such as X-ray, freeze electron microscopy, nuclear magnetic resonance and so on are used to analyze the protein structure, which can not keep up with the increase rate of amino acid sequence. This will cause massive samples to wait for months or even years in the laboratory. SWISS-PROT, an international authoritative database, shows that the accumulated protein sequence information has exceeded 560000. It is undoubtedly an "impossible task" to use traditional methods to complete this vast and varied sequencing, which requires a new approach. The wheels of scientific and technological innovation roll forward. As mentioned above, the development of artificial intelligence has brought opportunities for the efficiency of protein sequencing. Among them, the AlphaFold2 model enables AI to play an important role in the field of biomedicine. Generally speaking, if the prediction accuracy of artificial intelligence method exceeds 90 points, it can be considered that the prediction result is basically consistent with the protein structure obtained by the experimental method. The 92.4 score of AlphaFold2 can not only analyze the protein structure wedged into the cell membrane that is difficult to solve by X-ray crystallography, but also successfully solve the problem of protein folding, and even help researchers to manufacture proteins that do not exist in nature. These major breakthroughs brought about by AlphaFold2 mark that the basic theoretical research of AI assisted drugs has entered a new stage and set a new height for AI for Science. For example, almost all the drugs currently designed in the world act on proteins. Generally speaking, they need to be matched exactly like a key to unlock. The first step in this process is to determine which key opens which lock. In more professional terms, it is to find drug targets, that is, to figure out what kind of protein drug molecular action binds to. Through the artificial intelligence algorithm that can decode the protein structure, thousands of new drug targets can be quickly screened, thus greatly shortening the development cycle of new drugs; And the production of proteins that do not exist in nature can undoubtedly better help human beings cope with major challenges in an unprecedented way. End to end optimization of AlphaFold2, making AI more universal, AlphaFold2 provides a way for protein structure analysis and prediction, and opens a new window for the application of AI in biomedicine and other fields,