AI预测可以代替科学实验吗?——以AlphaFold破解蛋白质折叠难题为中心

Can AI Prediction Replace Scientific Experiments?:Focusing on AlphaFold Cracking the Protein Folding Puzzle

AlphaFold在蛋白质结构预测上的新进展代表着人工智能(artificial intelligence,AI)在生物医学领域的应用达到了一个新高度,但也引发了AI预测是否可以代替结构生物学家等问题。从人类破解蛋白质结构主要实验手段的发展史、AI预测蛋白质结构的科学原理、结构解析与生物学关系等角度分析,目前的AI蛋白质结构预测尚需要在实验结构生物学已取得的蛋白质结构的基础上进行训练,才能发展出预测未知蛋白质结构的能力。因此,AI预测目前还不能代替科学实验。但是,包括AI预测在内的AI应用将推进科学发现的进程。

The new progress of AlphaFold in protein structure prediction represents that the application of artificial intelligence in the field of biomedicine has reached a new level,but it also raises questions about whether AI prediction can replace structural biologists.This paper explores experimental methods history,the scientific principle of AI predicting protein structure,and the relationship between structure analysis and biological purpose,etc.The current AI protein structure prediction still needs to be trained on the basis of the known protein structure obtained in experimental structural biology,in order to obtain the prediction of the unknown protein structure.Therefore,AI prediction is not a substitute for scientific experiment currently.However,AI applications including AI prediction will advance the process of scientific discovery.