AI pharmaceuticals are maturing, and data sharing needs to be broken



In early 2023, when ChatGPT (a chatbot program developed by OpenAI in the United States, released on November 30, 2022) swept the world, AI pharmaceuticals, that is, artificial intelligence-driven drug research and development, also stood on the new outlet.

According to AI consulting firm Deep Pharma Intelligence, as of December 2022, the total investment of 800 AI pharmaceutical companies worldwide reached 5.93 billion US dollars, a 27-fold increase in nine years. In the first quarter of this year, there were more than 28 investments in AI pharmaceutical companies, with an average investment of $38 million.

On the map of AI pharmaceuticals in China, Shanghai Zhangjiang, the "old pharmaceutical valley", occupies a leading position, leading the country from the number of enterprises to the scale of pipelines. According to the publicly disclosed information statistics of Zhangjiang Group, there are a total of 99 enterprises in the country involved in the field of AI+ medicine, of which 34 are in Shanghai, and 25 Zhangjiang enterprises account for 25% in the country; There are 83 and 30 projects of AI+ pharmaceutical products in the preclinical research stage and clinical trial stage, respectively, while Zhangjiang accounts for 47% and 40% in the country, respectively.

In October 2021, under the initiative of Chen Kaixian, Jiang Hualiang, Rao Zi and three academicians of the Chinese Academy of Sciences, Zhangjiang AI New Drug Alliance came into being. By 2025, Zhangjiang Pharmaceutical Valley's "AI smart drug ecology" is expected to gather 300 active institutions, 30 innovation consortiums and 30 enabling platforms, and AI is expected to help add 30 new Class A new drug pipelines (pipelines, referring to a number of drugs in the development stage of pharmaceutical companies, including preclinical and clinical research) every year.

Today, the power of AI to accelerate target and drug discovery has been recognized by all, and the real test is still in the clinical stage. After Exscientia stopped developing the world's first AI-designed drug to enter clinical trials, the latest bad news is that Benevolent AI, another British AI pharmaceutical leader, recently announced that a drug candidate for the treatment of atopic dermatitis failed to meet a secondary efficacy endpoint in a phase II clinical trial.

In the future, which AI-assisted design or even designed from scratch drugs can be the first to successfully cross the "valley of death" of Phase II clinical trials, only time can give the answer. But there is no denying that as more and more pharmaceutical companies open their arms to AI, AI-enabled drug design is already unstoppable.

AI pharmaceutical "head on"

The biopharmaceutical industry has long been known as the "Double Ten" rule, that from the start of research and development of a new drug to the final approval of the market takes an average of 10 years, the investment cost of about $1 billion - many industry reports estimate the figure is several times more. In addition to the long cycle times and high costs, pharmaceuticals are a high-risk business, with industry estimates putting the global success rate of new drug discovery at between 2% and 15%.

Why is it so hard to develop new drugs? On the one hand, the human proteome, the difficult drug targets account for more than 75%, conventional targets are about to be developed, the track is particularly crowded; On the other hand, a drug candidate must meet a combination of conditions in multiple dimensions: solubility, activity/selectivity, toxicity, metabolism, pharmacokinetics/efficacy, and composibility.

Today, 60 percent of the disease has no effective drug, and 50 to 70 percent of patients do not respond to blockbuster drugs. A large number of clinical needs are not being met, and the industry urgently needs new drug development tools and paradigms, so AI has attracted the attention of a large number of entrepreneurs and investors.

Before 2012, the application of AI technology in drug research and development was still in the early stage of exploration, mainly including target identification, drug molecular design, virtual screening and so on. In the following five years, with the increasing maturity of machine learning, deep learning and other technologies, the advantages of AI application in drug discovery are expanding, and the scope of application is also extended to clinical trial design and prediction, "old drugs and new" optimization design scenarios.

2017 is regarded as the starting point of AI pharmaceutical industrialization. In September of that year, American AI pharmaceutical startup Atomwise announced that it had received $45 million in financing, becoming the largest financing in the AI pharmaceutical field. Iconic AI pharmaceutical companies, such as Exscientia in the UK and BenevolentAI in the United States, also made important breakthroughs in this year, and small molecule drug candidates developed by AI (small molecule drugs mainly refer to chemical synthetic drugs) began to emerge.

Insilico Medicine, headquartered in Hong Kong, China, is the first company in the world to explore the use of generative adversarial network (GAN) and generative reinforcement learning (RL) artificial intelligence technology for drug discovery, and has become one of the leaders in the field of AI pharmaceutical, leading drug research and development center in Shanghai is located in Zhangjiang. In February 2021, InSI announced that for the first time in the world, artificial intelligence was used to discover INS001-055, a drug candidate with a new target and a new molecular structure, for the treatment of idiopathic pulmonary fibrosis.

From the target discovery to the nomination of pre-clinical candidate compounds, InSI Intelligence only spent millions of dollars in research and development funds, which took 18 months, which attracted the attention of the industry. In March 2022, the drug was approved to enter phase I clinical trials in China, which is also the first AI drug to enter clinical trials in China. In January 2023, InSI released positive top-line data from its Phase I clinical trial. In February of the same year, InsiliconIntelligence announced that the drug candidate received orphan drug certification from the US Food and Drug Administration (FDA) and was about to start Phase II clinical trials, injecting more imagination into the prospect of AI pharmaceuticals.

A number of Zhangjiang enterprises interviewed by Surging technology said that AI pharmaceutical is not born out of nowhere, but the extension and evolution of traditional computer-aided Drug Design (CADD).

"With the improvement of deep learning technology, the concept of AI pharmaceutical has been put forward more and more specifically, and it turns out that AI is already used in many places." Chen Chunlin, rotating chairman of Zhangjiang AI New Drug Development Alliance and founder and CEO of Medisi, introduced that from the early random screening of candidate drug molecules, to the prediction of physical and chemical properties of small molecule compounds through computer simulation technology, and then to the rise of structural biology, based on the exploration of the interaction mechanism and binding site characteristics between drugs and proteins, The drug discovery and optimization process is beginning to become more predictable and controllable.

"But these are just changes above the point. Human experience can only be limited to one domain, for example, if a person is very good at synthesizing a particular sequence of compounds, another sequence of experience may be inaccurate." Chen Chunlin said that American chemist Christopher Lipinsk had proposed Five basic laws of drug design (" Rule of Five "), "now AI can consider these various factors, which is the biggest breakthrough brought by AI." Because there are more databases, although it's not the most comprehensive."

Fan Mengqi, vice president of the biomedical business group of Deep Trend Technology, said that AI can accelerate molecular virtual screening, help screen a larger chemical space in a shorter period of time, and help the discovery of seedling compounds. At the same time, the free energy perturbation (FEP) algorithm can further accelerate the lead compound optimization stage. "For example, under the traditional paradigm, the optimization stage may have to design and synthesize hundreds of molecules, waiting for several months of synthesis and biological experiment time, and now the designed molecules can first calculate the affinity with FEP, eliminate the molecules with insufficient activity, and finally only need to carry out biological experiments with the first synthetic activity, which can save a lot of time and cost."

Redefining "Structure is King"

In December 2018, AlphaFold, a protein three-dimensional structure prediction platform built by DeepMind (a British artificial intelligence company that was acquired by Google in 2014), showed amazing strength in the CASP13 competition (an international protein structure prediction competition), matching the structural accuracy of laboratory analysis. Less than two years later, AlphaFold2 set another record in the field. In July 2021, DeepMind unveiled a database of millions of predicted protein structures for use by scientists and researchers, marking an important milestone in AI for Science. At the end of July 2022, DeepMind again announced that AlphaFold DB had expanded from 1 million structures to more than 200 million structures, expanding by more than 200 times.

Zhong Wenge, chief scientific officer of Rueger Medicine, pointed out that the prediction accuracy of AI tools has a lot to do with known protein structures. The dynamic properties of protein conformation are particularly important for drug molecular design. "Changes in the binding pocket are key, but the more and more dynamic the protein, the more uncertain the prediction of the AI tool." Small deviations in specific amino acid residues, or even in combination with whether or not there are water molecules in the pocket, may be good enough for AI structural prediction, but for the drug hunter, it is the difference between whether the drug is active or not."

Fan Mengqi also believes that the structural accuracy of the current AI prediction is still a certain distance from the actual requirements of doing medicine, especially near the pocket, and the accuracy of the prediction is higher, so it is also necessary to combine dynamic simulation and other means to further refine it.

In May 2022, Deeppotential Technology released Uni-Mol, a universal molecular representation learning framework based on the three-dimensional structure of molecules, combining 200 million small molecule data and 3 million protein pocket data for common training. Fan Mengqi said that on the basis of this pre-trained model, only a small amount of data needs to be fine-tuned to achieve a prediction accuracy far beyond the previous model, and Uni-Mol has won the first place in 14 of 15 public test sets, and the molecular characterization ability is strong.

Deeppotential has also successfully applied Uni-Mol to more fields, such as material design. "Recently, we significantly upgraded Uni-Mol's model framework with the launch of Uni-Mol+ and won the top spot in the international authoritative academic competition OGB-LSC's prediction of quantum chemical properties. In the future, it will help Deeppotential's Hermite® drug computing design platform calculate more accurately." Fan Mengqi said.

According to him, traditional CADD is often based on traditional molecular dynamics (MD) sampling to simulate dynamic processes, but it is difficult to capture complete conformational changes when calculating a large protein system. The RiDYMO™ Enhanced dynamics platform created by Deeppotential uses AI to capture protein dynamic changes and explore the complete conformation space, enabling protein functional studies and allosteric/allosteric systems to find drug candidate molecules.

For example, he said, the company has a self-developed pipeline for tumor therapy, using enhanced kinetic methods to find budding compounds, "this target is an inherently disordered protein (IDP), in the human body lacks a stable three-dimensional conformation, the traditional CADD method can not effectively describe this type of protein, let alone rational development from scratch."

"We do not emphasize how fast the research and development time is, but really focus on the project logic of the project itself, especially for some difficult drug targets with great potential but not yet ready drugs, and it is not time to catch up." When you see something that no one else has seen yet, it's also a good way to develop it to a certain stage and then look for cooperation."

In Zhong Wenge's view, "AIDD (AI Drug Discovery) plays a great role in many aspects of innovative drug research and development, but its full realization may take a long time."

"Structural biology is the 'gene' of Regal medicine." Zhong Wenge said that Rui Ge Pharmaceutical is positioned as an innovative pharmaceutical company with "roots in China and global layout", and chooses to walk on two legs. One is to use the self-developed rCARD™ platform to carry out complex computer modeling with the help of tools such as computational biology, structural biology and computational chemistry. The second is the direct analysis of protein structure. At present, Ruege has worked with the Shanghai Institute of Pharmacology of the Chinese Academy of Sciences and the industry's top structural biology teams such as Weiya and Shuimu Future to analyze nearly 100 protein complex structures, including 5 cryo-electron microscope structures.

"Our self-created rCARD™ platform has been validated at an internal technical level." Zhong Wenge introduced that through man-machine combination, dry and wet test combination, Ruge Medicine spent ten months to discover the highly active and highly selective candidate drug RGT-419B molecule, achieving a complex balance of selective activity in the development of CDK4/2/6 inhibitors. At present, the Phase I clinical trial of RGT-419B is in progress.

"Our attitude is to embrace technology and embrace the future." Zhong Wenge said that Ruge Pharmaceutical has made efforts in two aspects: one is to transform the passive auxiliary thinking of CADD into active thinking, and the molecular optimization design is guided by information and calculation; The second is to actively adopt existing useful AI algorithms and collaborate with world-class AI teams to develop new methods. "We don't just do technology for technology's sake. We do everything with the goal of developing innovative drugs."

Domestic AI medicine market size and spatial forecast.

AI disenchantment, the industry to return to reason

"There have been a number of large and small companies competing to lay out the AI track, such as mushroomed momentum, but ultimately did not deliver satisfactory results to the market, that is, there is not much difference with the existing CADD." Recall Tang Qiusong, head of Roche China accelerator.

"Now everything is crowned with AI, including cell therapy is also crowned with AI, I think it is a bit exaggerated." It is now said that AI can design antibodies, proteins, RNA and so on, in fact, I think the most successful is small molecule drug design." Chen Chunlin believes that after a hundred years of research and development, small molecular compounds have a large number of literature records and method accumulation, which can provide a solid foundation for AI learning. Therefore, the easiest area for AI to achieve breakthroughs is to analyze and synthesize small molecular compounds to predict the toxic effects of drugs into the human body.

Medisi is a CRO company that actively embraces AI technology, and according to the company's internal statistics, in different cases, adding AI can indeed improve the speed and effectiveness of drug design, thereby improving the efficiency of pharmaceutical research and development. However, Chen Chunlin is still rationally and cautiously optimistic about AI pharmaceuticals.

He admitted that AI drug design is not omnipotent, and it is still impossible to get rid of human participation, "Sometimes the molecules designed by computers are difficult to synthesize, even insurmountable obstacles, such as some covalent bonds that cannot be synthesized." Although the probability of this happening is relatively small, it also shows that drug development can not be completely left to AI."

"Drug development is a process of trial and error, whether it is based on AI platform or traditional experimental drug development, it needs to be analyzed and iterated on the basis of multiple tests in order to screen out the optimal compound." Therefore, AI technology and traditional CRO can be said to jointly catalyze drug research and development in the speed, difficulty and breadth of mutualism and fusion promotion."

Chen Chunlin said with a smile that he originally wanted to promote Zhangjiang AI Pharmaceutical Alliance, precisely because CRO companies also have an important place in the AI pharmaceutical ecosystem: many AI technology companies do not have wet laboratories, and will choose to carry out experimental cooperation with CRO companies to verify algorithms.