"The application of AI (Artificial Intelligence) technology in drug research and development has attracted great attention from research institutes and the pharmaceutical industry. AI begins to empower the target discovery and confirmation of drug research and development, the discovery and optimization of drug lead compounds, drug pharmacokinetics and toxicity evaluation, and other stages, which will become one of the key core technologies of drug research and development in the future." On July 12, jianghualiang, an academician of the Chinese Academy of Sciences and a researcher of the Shanghai Institute of Materia Medica of the Chinese Academy of Sciences, made a keynote speech on the second issue of the 2022 "understanding the future" Science Lecture, "ai+ molecular simulation and drug research and development". In the lecture, Jiang Hualiang introduced the progress and trend of international innovative drug research and development. He believes that there are some key problems in small molecule drugs that need to be solved by AI. For example, the calculation speed of combined free energy has been increased by 3-5 times compared with the past, and only when the speed is increased to tens of thousands of times, the combination of free energy prediction accuracy and small molecule drug design efficiency is expected to achieve an essential breakthrough. According to Jiang Hualiang, the pain points of high investment and long-term in the pharmaceutical field are difficult to improve in a short time, but AI will have great potential in predicting the success rate of clinical candidate drugs. "In clinical trials, only one of about 10 candidate drugs will succeed, and we have accumulated clinical data of tens of thousands of drugs, including general data of a large number of failed drugs in previous clinical trials. Through modeling and calculation, we can predict and exclude the failed drugs in clinical candidates, and better target the drugs that may succeed." Jiang Hualiang said. Gao Yiqin, Professor of the school of chemistry and molecular engineering of Peking University and deputy director of the Department of science of Peking University, said in his lecture that the traditional molecular simulation is seriously limited by the space-time scale when applied to complex chemical and biological molecular systems. AI technology represented by deep learning can establish an organic connection between theory and calculation, theory and experiment, calculation and experiment, It has become an important tool to break through the bottleneck of traditional molecular simulation and empower molecular simulation and molecular science. According to reports, based on physical models, scientific experimental data and artificial intelligence algorithms, Gao Yiqin's team developed a number of molecular simulation methods combined with deep learning, and achieved excellent results in the global protein structure prediction competition (Cameo). However, the application of AI in drug research and development is still in its infancy. Jiang Hualiang said that it is necessary to develop new AI technologies for drug research and development, and closely combine them with traditional drug molecular design and experimental technologies, so as to truly empower drug research and development. Taking small molecule drug design as an example, Gao Yiqin mentioned that data is the biggest bottleneck restricting small molecule drug design. "At present, there are very few reliable data that can be really obtained, and there are still problems in data, such as inconsistent indicators and difficult access to sensitive data". Xiexiaoliang, the moderator of this lecture, director of the future forum and Professor Li Zhaoji of Peking University, also said that at present, enterprises have replaced large-scale small molecule drug screening with free energy computing, and microfluidic screening technology has also been used in experiments to increase flux, thereby significantly reducing costs. However, due to the insufficient amount of small molecule data and database, the machine learning prediction of small molecule drugs cannot be realized, which is a great challenge for small molecule drug design. Gao Yiqin believes that by integrating single-cell omics information and establishing a reliable cell response model, AI can make some predictions on the downstream of drug research and development. "If the flux is high enough, the cell model can be used to predict the small molecule entry into the membrane, protein signal transduction, protein nuclear transport, etc. in the design of macromolecular drugs and small molecule drugs. With the continuous self-learning and optimization of AI, the accuracy of prediction will gradually improve. If it is built into a public open platform, the whole pharmaceutical research and development will benefit." (outlook new era)