Recently, several major events have taken place in the AI pharmaceutical circle. Sanofi and exscientia, a British drug R & D AI technology service provider, reached a huge AI pharmaceutical order worth 33.1 billion yuan. Yingsi intelligence announced the discovery of new drugs twice within half a year and took the lead in entering a new stage of clinical trials. Aion labs, an AI drug research and development laboratory jointly established by pharmaceutical giants such as AstraZeneca, Merck, Pfizer and TIVA, announced its official launch. If 2020 is the first year of the outbreak of AI pharmaceutical and has achieved a breakthrough from "0" to "1", after two years of continuous warming, this track has entered a new stage from "1" to "10" by leaps and bounds. Many industry experts believe that AI pharmaceutical redefined the pharmaceutical process and brought a great change to the whole pharmaceutical industry. How does AI make drugs? For most ordinary people, biopharmaceutical is a strange field. In the movie "I'm not the God of medicine", the sense of helplessness of the anti-cancer group in the face of sky high price drugs makes the audience feel the same and cry countless tears. Through this film, the status quo of the pharmaceutical, drug production and drug sales industry chain, which has always been little known, has gradually emerged, allowing the public to see some of them. A person engaged in the pharmaceutical industry said: "as clinicians, we are also very helpless. The drug price is too expensive. Now many patients are taking purchasing drugs and even raw material powder." For pharmaceutical manufacturers, new drugs are expensive, which lies in R & D. The "double ten law" is a "big mountain" on their head, that is, it takes 10 years and 1 billion US dollars to successfully develop a new drug. Even so, only about 10% of new drugs can be approved into the clinical stage, and finally only a smaller proportion of new drugs can be put on the market. The huge R & D cost, long R & D cycle and unpredictable failure rate make the R & D of new drugs a thankless thing. Although pharmaceutical companies have been increasing investment for decades, the number of new drugs on the market invested $1 billion is reduced by half every nine years, a phenomenon known as "anti Moore's law". Generally speaking, the process of new drug discovery needs to determine the target of a disease first, and the target is equivalent to a "lock". Researchers need to design and screen the most appropriate molecule as a "key" among the possibilities of many drug molecules. Take the divine medicine "Gleevec" in "I am not the God of medicine" as an example, its target is a fusion protein, which can inhibit the fusion protein through drug small molecules to control the development of chronic myeloid leukemia. Whoever knows more about protein can find the "key" to crack major diseases and develop new drugs. "Traditional proteomic analysis techniques and methods are not completely suitable for studying protein systems. What it lacks is the process of quantitative data accumulation of proteins, and it does not have a suitable algorithm." Said Guo Tiannan, a distinguished researcher at West Lake University and founder of West Lake Oumi. The only option is to make the traditional drug discovery technology no longer appear on the stage. "The main role of AI is to simulate and calculate the mechanism of the combination of candidate drug molecules, compounds, proteins and genes. Typical application scenarios include virtual drug screening, protein structure prediction, etc." "This involves a large number of traditional high-performance computing applications such as molecular dynamics, quantum chemistry and quantum chromodynamics. Machine learning algorithms are also introduced in the process of learning and predicting the existing protein structure. It is basically a typical HPC (high-performance computing) + AI application scenario," he Wanqing, head of Alibaba cloud high-performance computing, told China Electronics News In short, the data-based artificial intelligence medicine R & D paradigm is essentially to summarize the drug R & D laws other than expert experience through machine autonomous learning data and mining data, so as to optimize all links of the whole process of drug R & D. It not only greatly improves the efficiency and success rate of drug R & D, but also effectively reduces the R & D cost and trial and error cost, so that the pharmaceutical industry can see the dawn of getting rid of the dilemma of "anti Moore's law". Who's on the track? Hitherto unknown, the covid-19 pneumonia epidemic has been an unprecedented focus on the biomedical industry. According to incomplete statistics, there were 77 financing events in the global AI pharmaceutical field in 2021, with a cumulative financing amount of US $4.564 billion (equivalent to about 29.073 billion yuan), an increase of 152% compared with 2020. 34 financing events occurred in China, with a total amount of more than 8 billion yuan. By the first quarter of 2022, there had been more than ten new financing events. Jingtai technology, yingsi intelligence, Shenshi technology and other "young" companies have made rapid progress and become the "pet" of capital. Founded in 2015, Jingtai technology has completed financing of nearly US $800 million (equivalent to about RMB 5.096 billion), setting a new record of financing amount in the global AI pharmaceutical field, with a total valuation of more than 13 billion yuan. Shenshi technology, founded in 2018, pioneered the new paradigm of "multi-scale modeling + Machine Learning + high-performance computing", and completed four rounds of financing in just 18 months. The core members of the team won the highest award "Gordon Bell Award" in the field of global computer high-performance computing in 2020, and their related work was elected as the top ten scientific and technological progress in China and the top ten technological breakthroughs in the field of global AI in 2020. Founded in 2014, yingsi intelligent has found a preclinical candidate compound with a new mechanism for the treatment of idiopathic pulmonary fibrosis (IPF) by using AI for the first time in the world. The whole research and development process took less than 18 months and about US $2 million, setting a new record of the speed and lowest cost of new drug research and development. At present, it has also completed six rounds of financing, with a cumulative amount of more than US $300 million. At the same time, traditional pharmaceutical enterprises and Internet giants have made efforts to invest and cooperate or play in person. For example, Fosun Pharmaceutical, together with silicon intelligence, has jointly promoted the research and development of AI drugs for multiple targets around the world. Yingsi intelligence received a down payment and milestone payment of US $13 million. This is the largest advance payment received by AI pharmaceutical enterprises in China so far. According to insiders, at present, pharmaceutical enterprises and AI pharmaceutical companies are mostly in the water testing stage. The cycle from early research to clinical stage is getting shorter and shorter, and pharmaceutical enterprises have a clearer understanding of the role of AI. Although large orders such as Fosun Pharmaceutical and yingsi intelligence are rare in China, more and more pharmaceutical enterprises are willing to test the water for AI pharmaceutical companies with some R & D directions, and the cooperation amount is also increasing year by year. Let's look at the artificial intelligence drug R & D and big data platform jointly developed by Alibaba cloud and ghddi, the artificial intelligence drug discovery platform "Yunshen intelligent medicine" released by Tencent, the propeller paddlehelix open source tool set launched by Baidu, and the medical agent eihealth established by Huawei... The rapid progress of these giants has greatly promoted the whole track to move forward in depth. "Through cooperation with biopharmaceutical enterprises, R & D institutions and technology companies in the field of service drug design, such as global health drug R & D center ghddi, Shenshi technology and Jingtai technology, Alibaba cloud high-performance computing deploys life science HPC applications and AI applications in parallel on the e-hpc platform by building a full link Life Science application solution based on the e-hpc elastic high-performance computing platform of cloud supercomputing, so as to achieve high-throughput and rapid development Fusion of row computing and AI algorithm. " He Wanqing said. "At present, Shenshi technology focuses on the two vertical fields of drug design and material design. In the field of drug design, we provide the business model of" software + joint research and development ". At present, the platform software has reached long-term business cooperation with dozens of academic and industry customers at home and abroad, and more than 10 pipelines in joint research and development are in the preclinical development stage." Sun Weijie, founder and CEO of Shenshi technology, told China Electronics News. It is reported that the two have also reached a strategic cooperative relationship. "The biggest advantage of Internet giants lies in computing power, algorithms and strong capital. The 'bottom card' of traditional pharmaceutical enterprises is a mature pharmaceutical research system and a huge data pool. Start-ups are more brave in exploring trial and error, which can bring a new perspective and innovative thinking. There is competition and mutual achievements among all parties. On the whole, the business model of AI pharmaceutical is gradually clear, and the ecosystem is also moving towards a virtuous circle." Expert analysis said. Industry spring is coming? In ancient times, Shennong tasted all kinds of herbs, and now AI technology has opened a new chapter in the pharmaceutical industry. From the concept to the development, the role of artificial intelligence technology in biopharmaceutical has developed from the initial computer-aided drug design to the current artificial intelligence drug research and development, and is even expected to run through the whole process from drug target discovery to early drug discovery in clinical trials. Dare to imagine that one day in the future, AI will be able to change from an auxiliary tool to a leading tool, and even independently shoulder the heavy responsibility of drug research and development? Although the imagination space is huge, the road ahead is not smooth. Globally, AI pharmaceuticals are in an early stage, which is far from mature technology and real market. Data, computing power, algorithms and talents are the "obstacles" that track players have to face. Sun Weijie said frankly: "on the one hand, most AI pharmaceuticals on the market are still at the stage of data-driven AI model assisted drug molecule discovery. At present, they are mainly faced with the problems of sparse total data and standard deviation of data, corresponding to the typical challenges such as ADME / T property prediction and molecular generation; on the other hand, we think that AI pharmaceuticals have another paradigm, that is, from physical model driven AI to traditional CADD (Computational assisted drug discovery) model innovation, the biggest challenge is that the computational efficiency and accuracy can not be balanced, such as the traditional molecular dynamics simulation is still unable to cope with complex protein systems and protein dynamic conformation sampling. " "For some time to come, AI pharmaceutical will continue to break through and evolve under the two paradigms of data-driven and model driven." Sun Weijie said, "the development of any paradigm essentially depends on whether AI or computing can really replace some experiments, so as to improve efficiency. We believe that in the next three to five years, the emphasis will not be on AI pharmacy, but on AI becoming a necessary means and standard configuration of drug discovery. At that time, we will no longer deliberately emphasize AI pharmacy, but will take it as a widely used method." At the same time, he Wanqing pointed out: "the biopharmaceutical field is currently a booming 'old tree and new bud'. The challenges to computing come from the complexity and scale of basic algorithms, algorithm innovation and other aspects. Therefore, the use of large-scale high-performance computing parallel cluster and GPU acceleration has become a universal first choice." He believes that biopharmaceutical is facing the integration of the medical knowledge of the whole mankind. Therefore, the unique connectivity and elasticity of cloud computing can help break the isolated island of research and development and promote the reuse and innovation of data results. In addition, for AI pharmaceutical, a multi-disciplinary and high barrier industry, the scarcity of talents may be a more fundamental pain point. Duan Hongliang, a doctor of medicine at the Shanghai Institute of pharmacy, Chinese Academy of Sciences, said bluntly: "only when you have a deep understanding of both pharmacy and AI can you know which key problems in the pharmaceutical link AI is good at solving. Only when you find the coincidence point between the two can you solve them