AI application revelation: Why do capital love drug research and development?

For AI practitioners, it is probably not a good year in 2022.

The capital of capital is getting higher and rising, and the market value of unicorns such as commercial soup and graffiti has shrunk significantly. After the pessimism of the secondary market was transmitted to the first -level market, the financing frequency of the entire industry fell cliff -like. The once -hot capital darlings were encountering "the dilemma of innovators".

There are many sounds of singing AI. There are two most common reasons: one is that there are too few commercialization and maturity at this stage, and the other is that the profitability of AI unicorn is too weak. The logic of "burning money for growth", which has been tried and tested in the past, has been difficult to make interest in the capital market.

When the external environment is sorrowful, AI pharmaceuticals have become one of the few exceptions. This emerging industry that only entered the public vision in 2020, standing against the trend of capital, some companies that established one or two years have received 100 million yuan in financing, and some industries have even won tens of billions of orders.

Why is AI being "abandoned" by capital? The research and development of drugs is often favored by the inside and outside of the circle. It is clear that these issues can not only figure out the reality of AI pharmaceuticals, but also find some inspiration for the application of AI.

01 A natural AI scene

The popularity of AI Pharmaceuticals is really a bit of clever factors.

At the 14th international protein structure prediction contest held in 2020, Google's Alphafold2 successfully predicted the three -dimensional structure of the protein based on the gene sequence. Under the tension of the black swan manufacturing of the epidemic, such a initiative quickly set off a boom in the world.

At present, almost all drugs are acting on protein. The principle is to change the function of protein through the interaction of drugs and target proteins, and achieve the effect of treatment, just like the relationship between locks and keys. If you can accurately predict the three -dimensional structure of the protein, you can accurately design the corresponding key.

However, the formation of protein requires the translation of DNA to RNA and RNA to translate into amino acid chains, and then synthesize macromolecular proteins by amino acid dehydration. There are 22 amino acids that make up proteins. One protein contains tens to tens of thousands of amino acid chains, and has a complex three -dimensional structure, so that many scientists can fully explain a certain protein structure in their lives. And metaphor.

It is from 2020 that investment institutions have teach AI pharmaceuticals. According to the "Artificial Intelligence Index" report released by Stanford University, funds invested in the field of AI drug research and development in 2020 increased to $ 13.8 billion, exceeding 4.5 times the same period last year; Doubles, the total financing has increased by about 10 times year -on -year.

The performance of capital is not crazy, but it is not blind behind the crazy, but to see the attractive "money scene" contained in AI Pharmaceuticals.

The drug research and development industry has a well -known "double ten law", that is, 10 years and $ 1 billion, it is possible to develop a new drug. According to the statistics of Nature, the "Double Ten Law" is actually an ideal form. In reality, a new medicine takes an average of 10 to 15 years from R & D to approval, which takes about $ 2.6 billion, and clinical success rate Less than 10%. At the same time, the terrible "anti -Moore law" also appeared. Even if the pharmaceutical company's R & D expenditure has increased in the past few decades, the number of new drugs in exchange for $ 1 billion will be reduced by half every 9 years.

The long cycle, high cost, and low power of the new drug research and development have undoubtedly left a huge place for AI for AI: through the self -learning data of the machine, digging data, summarizing the rules of drug research and development of expert experience, and then optimizing the drug research and development process In all links, it can not only improve the efficiency and success rate of drug research and development, but also reduce R & D costs and trial and error costs, and drive the industry to get out of the shadow of "anti -Moore's law".

02 The current situation of AI Pharmaceuticals

As for whether the AI ​​pharmaceutical is a new bubble, you must also find the answer from the logic.

From research and development to mass production, it involves more than a dozen processes including target discovery, compound synthesis, compound screening, crystal prediction, pharmacological assessment, clinical design, drug redirection, approval production, etc., covering drug discovery, pre -clinical research , Clinical research, approval and listing: four stages. The long cycle and long process of drug development provides more entry points for AI than other fields.

It is only different from the high enthusiasm of overseas pharmaceutical companies. At present, only a few pharmaceutical companies such as Yin Kangde and Chinese biopharmaceuticals are actively involved. The protagonists of the market are actually technology giants and entrepreneurs.

The representatives of the former include Huawei, Tencent, Ali, Baidu, etc. The giants often do not participate in drug research and development, but focus on the required algorithms and computing power. For example, Tencent's cloud -Shenzhen pharmaceutical, the positioning when launching is a drug discovery platform, providing algorithms, databases, cloud computing and other services, target customers are pharmaceutical companies; Baidu's Bai Tu Shengke, also the output algorithm model and computing power basis, inheritance The platform model that the Internet is good at.

The latter seems to be the main force of AI Pharmaceuticals. Typical examples include British Silicon Smart, Jingtai Technology, and so on. Among them, Yingsilia Intelligent has established an AI platform that integrates innovative targets, innovative molecular discovery, and clinical research results, including target discovery and multiple sets of data analysis engines, molecular design engines, clinical test results prediction engines and other components; crystal crystals; Jingjing Thai Technology has established a research and development system including intelligent computing, automation experiments, and expert experience. Under the exploration of giants and entrepreneurialists, domestic AI pharmaceuticals are not lacking of valuable results. For example, Baidu's Linearfold algorithm, shorten the whole genome of the new crown virus from 55 minutes to 27 seconds; Yingsilia Smart announced in February 2021 that it only took 18 months and $ 2.6 million to invest, and then it was invested. The new target of special pulmonary fibrosis has been developed, saving a lot of drug discovery costs.

Photo source: Euou Think Tank

There are no shortage of these results, but it is difficult to hide such a fact: the layout of the giants is still at a strategic level, and the large -scale landing application still takes time to verify. It is mainly based on drug discovery, which may greatly reduce the cost and time of drug discovery, and the impact of total cost of drug research and development can still be very limited.

In other words, AI Pharmaceuticals is an industry that has not been fully verified. Natural language processing, image recognition, deep learning, cognitive calculation and other cutting -edge technologies have gradually landed, but the depth of landing may be far less than the security industry. The reason for this is inseparable from the industry's attributes of drug research and development. After all, the target is determined by the clinical trials. It takes only two or three years of the AI ​​pharmaceutical wind, but it cannot hinder the emergence of questioning.

03 Commercialization of self -made factions

The most harsh question is: No AI drugs are listed.

Such a sound is mixed with too many subjective colors. AI plays only auxiliary role in the pharmaceutical process. It is impossible to go around the industry's inherent processes and mechanisms. It is impossible to complete ten years in two or three years. However, it also reveals a established fact that most of the AI ​​pharmaceutical distance is still far from commercialization. Pharmaceutical companies only have a limited taste of new technologies, and AI is still a bonus.

Any pessimistic signal is transmitted to the market, which will affect some people's decisions, and AI pharmaceuticals cannot escape the curse of being sang. Jingtai Technology is rumored to want to move to the Hong Kong Stock Exchange after listing in the United States, and eventually listed on A shares. The market value has doubled by nearly double the previous valuation; The market value of the US dollar has dropped sharply to $ 2 billion.

Temporary twists and turns are not to let AI pharmaceutical fall into the altar. Compared with the homogeneity of security and quality inspection, AI pharmaceuticals have evolved from shallow to deep into three business models:

The first is to build the AI ​​technology platform. It is hard power such as computing power, data, algorithms, etc., mainly to collect authorized use fees from customers such as pharmaceutical companies, and it is also the main battlefield of science and technology giants; the second is to help pharmaceutical companies or CRO companies complete the research and development tasks For example, the appropriate compound is screened according to the established target; the third is the self -built laboratory and the R & D pipeline.

The revenue path of the technical platform is not yet to be repeated, and the point is the feasibility of the latter two business models. Due to the uncertainty of the pharmaceutical medicine, the payment methods such as down payment and milestone price have been derived. At present, the average down payment of domestic pipelines is 2.8 million US dollars. The milestone price may reach tens of billions of yuan. Relying on the funds of pharmaceutical companies by the cooperation pipeline to avoid the situation of "lack of food". The risk of self -developed pipelines is relatively large, but in the investment list of such enterprises, the head of the head pharmaceutical company can be seen.

The middle fate is not complicated. Compared with traditional pharmaceutical giants, AI pharmaceutical companies are often not large. Even if there are extremely high risk costs, the "cost of failure" is easily accepted by the industry. An ideal format is that the early stages of AI pharmaceuticals were promoted by the capital market. After the prospects and feasibility of the market were verified, the pharmaceutical companies bought a marathon -style long -distance running. During the period, some companies may be eliminated. forward.

At least for the time being, AI Pharmaceutical has shown good application prospects in the scenes of disease mechanisms and target research, target drug design, compound screening, crystal prediction, and preclinical auxiliary research. Drug discovery, as the cornerstone of the entire drug research and development process, is also the basis for the gradual step by AI pharmaceutical.

04 Paradigm innovation in performed

For example, the current AI pharmaceutical is at dawn.

Because of the complexity of biology and the lack of clinical databases, AI's penetration of drug development stays in the front -end drug discovery link. In addition to the "criticism" of no difference, there seems to be another explanation: the dawn of AI has been illuminated in the pharmaceutical field. It is not only efficient and cost, but also reconstruct the industry's collaboration paradigm.

In the traditional drug research and development process, medicinal experts often need to propose 5000-10,000 compounds for drug screening according to experience, screen out about 250 compounds to enter the clinical study, and then find 5-10 compounds for clinical trials. In the end, there is one. The two compounds pass through clinical testing, which is tantamount to "the sea fishing."

The rise of AI pharmaceuticals is not long, but many pharmaceutical companies are already trying to transform the process. For example, the introduction of neural networks during the filtering phase of the compound can screen more than 100 million compounds in a few days. According to the predictive score of the algorithm model, the compounds are ranked to list several to dozens of most likely compounds. Simply put, the principle of AI participating in drug screening is to use inductive reasoning capabilities to accelerate the screening of compounds, get rid of the deep dependence of the experience and knowledge of medicinalized experts, and become more scientific in the link. At the same time, the attitude of the pharmaceutical companies' foreign cooperation.

As early as 2019, 10 pharmaceutical companies including Johnson & Johnson, Roche, Sanofi, and Takeda participated in the famous Melloddy Alliance to share drug data through blockchain and federal learning; , Takeda, Lilly, and GSK participated in the ACCUMUMULUS SYNERGY in 2020, trying to enhance collaboration and data sharing in all regions of the world; Astraikon, Merck, Pfizer, Tiri, and AWS were established under the advocacy of the Israeli Biotechnology Fund AION LABS, AION Labs, co -developed AI technology and incubated new companies ...

Although the Melloddy Alliance has not yet reported the news of the three -year cooperation, it has not yet been reported that the change of the contract for renewal is an indisputable fact that the experimental data that was once regarded as the core business secret has already been in the core business secret. Sharing within a limited range, and the underlying logic of AI pharmaceuticals is to train with a large amount of data. The more sufficient the amount of data, the more accurate the prediction results of the algorithm.

The truth hidden behind these phenomena is that in the face of new species such as AI, pharmaceutical giants have not played the role of "conservatives" like many industries in many industries, but are actively promoting the application of AI, even if they are actively promoting the application of AI. Even At this stage, the implementation of the ground is relatively limited, but it is not ruled out that the possibility of AI application in clinical trials and other core links.

The relatively optimistic market environment may be the direct benefit of AI pharmaceuticals, and it provides a reference direction for domestic AI Pharmaceuticals.

05 Write at the end

Let's answer the question at the beginning of the article, the answer is very clear.

The rise of AI Pharmaceuticals is by no means accidental. There is a huge application potential in the industry itself, and it has achieved immediate results. In the specific landing process, the layout of different camps is a bit heavy, but the entire industry does not fall into the trap of the inner roll. There is a relatively rational division of labor; the most important point is that AI pharmaceuticals are not the story of "juvenile slaughter of evil dragons". Whether it is an entrepreneurial or pharmaceutical company, it is promoting the industry.

Perhaps at this stage, AI pharmaceuticals have irrational prosperity. Some of the players on the field are speculators. They are still in the starting stage of the market. Many difficult problems cannot be resolved, but they also have the scarce quality of the industry, that is, the value of AI, the value of AI Never be blindly advocated. Most participants are trying to find the scenes of AI and pharmaceutical fit, so that the outside world can see a gradual progress and surprise.

- END -