Wen Shuhao: Use a good AI tools for the pharmaceutical industry open source.

Read: AI tools are increasingly being used in drug discovery!---- Last July, a team at Flinders University in Australia announced that a new seasonal influenza vaccine developed using AI technology had entered clinical trialsThe product is based on a set of ligand search algorithms (SAM), and the public title "first AI-designed drug" is more or less transmitting the industry's expectations for AI-empowered drug developmentIt was also noted that two months later, Canada's Deep Genomics also announced that "the first therapeutic candidate drug discovered by AI" had been successfully releasedThe compound, called DG12P1, took only 18 months thanks to AI's guidance from target discovery to pilot compound screeningAI tools are increasingly being used in drug discoveryWhile some key players point out that these tools offer significant opportunities for change in drug development, many in the industry are skeptical"It's normal to argue that any new technology will face this challenge before it becomes available"At the end of June, I interviewed Wen Shuhao, co-founder of Jingtai Technology, to respond to this disagreement, "AI certainly has its expertise, such as the processing of data, but there is room for improvement." Instead of pure technology, we should value the consequences of technology in drug development, and from this point on, digitalization, including AI, actually provides a new perspective on things"DrWen Shuhao, co-founder and chairman of Jingtai Technology, from macro to micro, what is the drug? Long-cycle, high-cost characteristics, has been a significant label in the new drug research and development industryIt is generally assumed that the successful development of a new drug would require a "$1 billion, 10-year" premise, but the reality seems to be becoming more complexAccording to a 2014 Tufts Center report, the funding threshold for developing a drug has been raised to $2,588 millionSpecifically, the cost of obtaining a new compound rose sharply after the drug companies actually spent $1,395 million, plus a nominal interest rate of 12-14% (cost of capital), which made up for capital time and opportunity costsIn terms of time, this changing trend makes drug development more difficultThe reason for this is related to the declining success rate of new drug development, increasing clinical trial costs, and more stringent regulatory conditions, while pharmaceutical companies are rapidly consuming new drug targets that have been discovered in the pastContinued growth in capitalized research and development costs "Industry may look at this more from a biological and chemical perspective, for example, how to design better clinical trial sourcing, verify the effectiveness and safety of a drug in the shortest time possible, and then drive subsequent listings." But based on the physical context, we want to understand the interaction of drugs and targets from a more microscopic perspective In fact, this kind of thinking is throughout the drug development process, from the early drug design to the subsequent clinical research, a clearer understanding of the drug, can help improve the success rate, reduce the cost of trial and error Wen explained Drugs are made up of atoms and electrons -- and this is also Wen Shuhao's entry point in drug development In 2014, Wen Shuhao and Jian Ma and Lai Lipeng were also studying physics at the Massachusetts Institute of Technology Influenced by the local entrepreneurial atmosphere, the trio decided to start from the drug crystal type they are good at, set up Jingtai Technology, hoping to accelerate drug research and development, improve the accessibility of innovative drugs The idea was endorsed by Pfizer, which also published a special article on quantum physics and how AI affects drug discovery and development: "In recent years, scientists have begun to use computer modeling techniques called crystal-type prediction (CSPs) for virtual crystallography By applying quantum physics, scientists can predict the behavior of electrons in molecules to determine their 3-D structures "The outside world often uses crystal prediction as a label for Crystal Technology, and in fact, as mentioned in the article, Pfizer scientists have detailed the huge potential of Crystal Technology's potential for application in the discovery of new drugs in the underlying physics, artificial intelligence, and cross-cloud computing capabilities Driven by these multinational drug companies, Jingtai Technology has created a new generation of drug discovery platform ID4, which is "physical and AI-super-computing" What's more, the platform has made several milestones in nearly 20 new drug development projects in China and the United States The discovery of new CSP and AI drugs involves many complex mathematical calculations, relying on the corresponding computing power "In the past, it was thought that the interaction between the drug electron cloud and the human target protein electron cloud was fixed and actually interfered with each other Wen Shuhao believes that although the energy of change is small, it takes a lot of calculation to calculate the results Due to the lack of past conditions, pharmaceutical companies usually abandon this link, resulting in poor crystal prediction, as well as the production of false positive molecules in the design of new drugs With the development of AI and cloud computing technology, these problems can be effectively solved Another notable phenomenon is the growing push by a growing number of big international drug makers to create chief digital officers This also largely reveals that the drug development industry is embracing a wave of multidisciplinary cross-cutting While the current compound talent gap is still large, in Wen's eyes, these changes at least reverse the traditional perception of drugs and allow them to explore a more micro-world The "three mountains" of AI drug development", artificial intelligence, not only brings new challenges to relevant scientists, but also to the biopharmaceutical industry and its established process of discovering and developing new drugs In December 2019, a paper published in Nature Review's Drug Discovery rethinks drug design in the AI era In drug discovery, the process of raising the characteristics of the pilot molecule to the candidate drug level is referred to as the "design-manufacturing-test-analysis" (DMTA) cycle This classic hypothesis-based approach first uses available data to make assumptions and design molecules (or select existing molecules from the library), then synthesize design compounds and test and validate assumptions with appropriate active testing This knowledge is then analyzed and translated into the basis of the design assumptions for the next cycle The advantage of AI is that, instead of relying on a large library of screening compounds, small amounts of compounds can be synthesized in each operation, with only the amount required for testing In other words, AI offers a range of options for improving DMTA efficiency By providing improved synthesis routes and optimizing reaction conditions, The AI algorithm model allows chemists to follow the most effective route, thus shortening the "manufacturing" phase It should be noted that a standard set of data sets is critical to the establishment of the AI prediction model Even complex algorithms cannot produce useful results if you make predictions based on limited data and limited understanding In fact, the data issue is one of the sources of disagreement in the industry about the role and size of AI's use in drug development The two cases cited in the first paragraph face the same challenge: some internal data is not available, AI data sets and training sets are limited, and quality issues affect the credibility of the results due to inconsistent data specifications However, Mr Wen has reservations: "Data cannot be ignored, but understanding of the data itself is equally important." At first, the size of the training set wasn't necessarily that critical "Take The experience of Jingtai Technology as an example, because the early did not have the conditions to obtain data from large pharmaceutical companies, Crystal Technology sacrificed huge computing power, through virtual data self-building capacity, to solve the most basic modules in drug design - this digital twin technology, can be mapped in the virtual space, thus reflecting the corresponding physical equipment life cycle process." On the other hand, Jingtai Technology in the specific project process, will also derive a series of available data Moreover, feedback on this data acts directly on the algorithm icing model, helping the latter to optimize the adjustment Interestingly, Wen Shuhao seems to care more about the available space of AI algorithms than the outside world "In each new drug development project, we evaluate how much the algorithm plays in each step, and even establish a corresponding quantitative standard Wen added Compared to the concept of "AI New Drug Discovery", Wen would prefer to see the process as "an efficiency upgrade in the pharmaceutical industry of digitalization and artificial intelligence" The connotation of upgrading, in addition to data acquisition and algorithm utilization of the change, but also includes a huge increase in computing power Microdrug molecular studies deep into the atomic electron level "In the microscopic way, they (Crystal Technology) are calculating the effects of electron effects on drug effects in drug molecules But because molecules have many moving and changing electrons, multiple calculations must be performed simultaneously, and the final answer may take billions of calculations Bruno Hancock, global head of materials science at Pfizer's Groton Research Station, noted in the previous article on Jingtai's scientific and technological cooperation that when cloud service providers calculate for these projects, Crystal Technologies became its largest user At present, Jingtai Technology is still using cloud service providers for computing But ideally, Wen also envisions building an in-house cloud architecture This can further reduce costs and speed up the project cycle, while at the same time, the algorithm of Crystal Technology can be integrated into the realization of synergies For the pharmaceutical industry open source to mass, as one of the earliest layout of domestic AI drug development technology companies, the growth of Jingtai technology is obvious to all Chinese At the end of 2015, Less than a year after its establishment, Jingtai Technology received 24 million yuan of Round A financing from Tencent; "AI, cloud computing and pharmaceutical technology companies", which was previously Wen Shuhao to Jingtai technology under the label According to its planned development logic, Jingtai Technology by algorithms to connect AI and drug research and development, which provides the possibility for funds in the field of TMT What's more, with the introduction of these investors, Jingtai Technology has been able to mobilize the latter's enormous computing power, and thus achieve the drug development process to promote This is back to the starting point of Jingtai technology "Our ultimate goal is to help the pharmaceutical industry open source, and one of the things that open source is that people can use this platform to generate more drug assets Wen Shuhao believes, "For large pharmaceutical companies, it means a new product pipeline, for small and medium-sized pharmaceutical companies is an opportunity to hatch and grow, for CRO companies to increase the corresponding research and development needs to bring them more business." "