Opportunity Of "Overtaking On The Curve" Of Chinese Pharmaceutical Industry: Computational Empowerment Leaps Over The "Dividend Crisis" Of Biochemical Technology

"Why is imatinib so expensive? The movie "I am not the God of medicine" tells the story of a group of good people who risk crime because of the high price of imatinib, and even startles the prime minister to personally approve the inclusion of him in medical insurance. But in fact, there is a reason why the price of pharmaceutical companies is expensive, because the research and development of drugs is too slow. It took 42 years, and there were several abortions in the research and development process. The cost of investment is immeasurable, and the high price is inevitable. " Recently, at an industry conference, Fang Xiangdong, a researcher at Beijing Institute of genomics, Chinese Academy of Sciences, pointed out to the media, including the 21st century economic report, that shortening the drug R & D cycle is particularly critical.

At present, a view formed in the industry is that computing can maximize the drug benefits and greatly improve the R & D efficiency of the biomedical industry.

Professor Tan Guangming, director of the high performance center of the Institute of computing technology of the Chinese Academy of Sciences, gave an example to the reporter of the 21st century economic report that since the advent of Google alphafold2 algorithm, not only can the test time be shortened from a few years to a few hours, but the investment cost has also dropped to tens of thousands of dollars, and the accuracy rate is no less than that of traditional experiments. In short, computers can help reduce the cost of trial and error, reducing 10000 biological experiments to 100 or 10 times.

The essence of computational medicine is to adopt the research paradigm driven by intensive data, with artificial intelligence as the method and supercomputing as the support. It can always look at life systematically and globally. To explore the life law and basic mechanism contained in data and knowledge by quantitative method, and provide services for biology and medicine with engineering method.

Tan Guangming believes that if the data of phase I and II clinical trials are good, the phase III clinical trials may still fail. We must have new technology to break the inherent "ceiling". Embracing new technology may fail, but it is also most likely to succeed. The efficiency of the pharmaceutical industry driven by biochemical technology is declining. A potential trend is to build a new technology system to push drugs from experimental driven to digital driven, and to promote the supply side to provide efficient and high-quality products.

The dilemma of medicine and medicine

At the above industry meeting, Zhang Chunming, executive vice president of the Western Institute of advanced technology, Institute of computing, Chinese Academy of Sciences, pointed out that there are three challenges in the field of Medicine: first, the technology dividend of target discovery has ended, and the new technology has not yet worked; second, clinical trials are the "valley of death", which is difficult to cross; third, the effectiveness of clinical drug use for diseases is poor, and market differentiation is difficult to compete.

Among them, the difficulty in differentiated competition is due to the serious homogenization of Chinese pharmaceutical products. The commercialization of many drugs in the United States will succeed if they are successfully developed. However, many drugs in China have been developed and commercialized in failure. Therefore, it is necessary to call for industrial innovation.

Business is not necessarily successful when it is academically recognized. Chen Bin, general manager of Huaxia Fund holding subsidiary and doctor of internal medicine of Tsinghua University, believes that problems in the scientific highland can be transformed into some industrial problems and integrated into the industrial value chain, which is the prerequisite for commercial success.

In recent years, although the number of domestic biopharmaceuticals approved has increased year by year, the degree of innovation is still insufficient. The research and development of innovative drugs are still focused on hot targets with serious homogenization, and the track is very crowded. Some popular targets, such as PD-1, PD-L1, Btk, etc., are being done by many enterprises. Even more than 100 enterprises are crowded in one track, so there is a certain blindness.

Zhao Yu, deputy director of Zheyuan Turing Darwin laboratory, Institute of computing, Chinese Academy of Sciences, believes that big data is not a tool for some pharmaceutical enterprises, but a quagmire. "Although the production of data has been industrialized, the ability to understand and use data is still very poor, especially in the field of biology and medicine."

There will be faults in the application of multi-dimensional and heterogeneous omics data to clinical practice, which is mainly reflected in two aspects. Firstly, clinicians can not understand the data and will not use data to help patients in diagnosis and treatment. In addition, the published high-level articles can not effectively solve the problems encountered in clinical practice. Fang Xiangdong said that building a knowledge network is the process of expressing scientific discoveries in a language that clinicians and downstream developers can understand.

Data production and understanding are totally out of balance, so the development of life science and medicine needs new technology engine. The pharmaceutical industry has always been supported by biochemical technology. Along this technical route, China's pharmaceutical industry can not surpass that of developed countries in Europe and America. Only computing technology can provide new opportunities and realize overtaking on curves.

Zhang Chunming said that looking at the biomedical industry and research, the dividend of biochemical technology has come to an end, and there is a lack of new technology platform. Computational medicine can provide medicine with a new technology system besides biochemical technology, and move jobs to computers. Computers do 80% of the work and people do 20%. Computational medicine platform can not only improve the success rate, but also reduce the experimental scale and save costs. According to the evaluation of major research institutions, it can save nearly 50 billion US dollars of research and development funds for new drug research and development, and truly realize the halving of drug research and development time and investment, and double the clinical efficiency. So the combination of computer and biochemistry is invincible.

Technology system of computational medicine

Although there are more than 200 individual cell mutations in a non-small cell lung cancer tumor cell, there are few known and helpful for clinical treatment, and no more than 10 of them can be used as markers in clinical practice. Although the other 200 variants are necessary for tumor growth and function, they do not play any role in the interpretation because they are not understood.

The use of data model can effectively improve the understanding of it. Niu Gang, director of the Zheyuan Turing Darwin laboratory, Institute of computing, Chinese Academy of Sciences, and doctor of cell biology and systems biology, said that gene mutation will cause certain consequences in cells, which can be measured and analyzed. Therefore, it is not necessary to build various data models and deep learning frameworks to simulate real cells without any biological knowledge, and further through knowledge Modules interpret any real cell function.

The human and disease-related functional data, which are re interpreted according to cell function, are called baseline data. Niu Gang said that only based on the baseline data can we do a good job in the whole process management of patients in advance. After getting the functional information of lung adenocarcinoma, we should first judge the metastasis direction, then judge the prognosis of the tumor (e.g. predicting the OS of the patient), and finally judge whether the treatment time is appropriate (such as the use of anti vascular drugs) and whether the immune drugs are applicable. If subclonal evolution occurs, we need to know its evolutionary direction.

Zhang Chunming also said that from genes to cell behavior, and then to disease manifestations, mathematical methods can be used to establish a link, and in this way, the effectiveness of clinical treatment can be judged and explained, and even the most suitable population can be selected for experiments to achieve the best drug effect.

The problem of computational medicine is to use human knowledge in production practice and scientific research. In the field of drug research and development, knowledge mapping can be used to integrate all human knowledge to write a "book". For example, AI has written a book "everything about autism", which can systematically sort out and summarize more than 56000 literatures published by human researchers. Knowledge map is a consensus concept in IT field, but domain specific knowledge map tests fusion ability.

There are more than 30 million articles in biomedical database. The basic assumption of computational medicine is to make full use of them, but how to use them becomes a difficulty. According to Niu Gang, the first step is to grasp the most core concepts related to the research content. The process of grasping is to rearrange the distribution of knowledge. As the knowledge piles up due to conformity psychology, the virtual height should be suppressed. The second step is to reconstruct the core content. AI needs to be carried out without any prior knowledge, so it needs to go through several rounds of iteration to get the really relevant content. The third step is to extract knowledge particles by classification. Each category represents a specific direction or research hotspot in this field. Finally, relevant databases are used to extract gene interactions, signaling pathways, drugs and other annotation information, so that the incremental knowledge and data can be added to every field of human cognition.

Zhang Chunming said that in the early stage of the global pandemic of new coronary pneumonia, when there was no large number of relevant literature reports, more than 14000 articles on coronavirus were mined by using knowledge mapping technology. Two conclusions were drawn through combing: one is that losartan, a blood pressure lowering drug, can prevent the critical illness of patients with new coronary heart disease; the other is that C21, a small molecule drug, can be used as a potential drug for the treatment of new crown pneumonia 。 These two conclusions were the first to be listed as the top ten progress of heart disease by American AHA last year, and the small molecule drug C21 was registered by a British pharmaceutical company and entered the second phase of clinical practice, and the current effect is ideal.

Computational medicine needs super high performance computing support. Professor Tan Guangming said that since there is no general software, general knowledge and no general algorithm, supercomputing is needed as support. However, supercomputing is easy to buy and difficult to use, which involves "parallel optimization technology". As the strategic layout of the Institute, "biomedical big data" has been studied for 20 years, which can process massive data quickly and efficiently.

In the field of application, Zhang Chunming said that different combinations of models can be used to build application-oriented interface tools, and tools can be added according to different needs to build digital experimental scenes of drugs, from computing target, computing structure to human drug matching, so as to truly use the platform to solve problems.

Computational medicine empowers pharmaceutical industry

One drug is not suitable for all patients, so it is very important to select the most suitable population by using computational medicine platform, and to improve the success rate of drugs through differentiated competition. Zhang Chunming said that the computational medicine platform should fully cover the above scenarios. A small amount of clinical trial data have been put on the platform. The most suitable population for drugs can be selected through data analysis. The typical representative to ensure the success of drugs by selecting the dominant population is IRESSA.

Originally, drugs can be sold to 10 people. After screening, they are likely to be sold to only two or more people. This may make pharmaceutical companies feel contradictory. However, there are three main advantages in selecting the most suitable people by using the platform. Zhang Chunming pointed out that the first is that it can be certified as a breakthrough therapy to shorten the time for medical insurance; the second is that the dominant population is clear, so the drugs can be priced differently; the third is to make the drug more effective on indications and help R & D enterprises improve their internal skills.

According to the data of clinical trials, we can infer the characteristics of the drug dominant population and infer which diseases the drug is effective against.

The success of a drug research and development is not easy. 10% of the drugs on the market can expand the indications, and hundreds of millions of people can be expanded through the platform, which can greatly increase the value of the drug. Pfizer, Novartis and Eli Lilly have demonstrated great commercial value in hormone positive HER2 negative breast cancer. More than 100 groups of new indications have been explored in the world, but the cost is expensive and the clinical trials are protracted. The digital drug research and development platform under the guidance of computational medicine has helped CDK4 / 6 inhibitors "calculate" a number of new indications. One of the new indications points to chordoma, a rare disease without drugs in the world. In clinical practice, after three weeks of single drug treatment, patients with recurrent chordoma who have failed in both manual and radiotherapy treatment according to AI judgment are suitable, The tumor was reduced by 37%. The other new indications are non rare tumors, which brings a broad imagination for global drug research and development decisions.

What's more, there are a large number of failed innovative drugs every month in the world. Unknown targets can be found through computational medicine platform, new industrial clusters can be created, and the value of clinical failed drugs can be rebuilt.

Zhang Chunming also stressed that the digital pharmaceutical laboratory of computational medicine will not replace the pharmaceutical enterprises, that is, 80% of them will work on computers and 20% will have to be done by people. However, enterprises that use computational medicine technology will certainly replace those enterprises that do not use computational medicine technology, and the trend is like this.

?