Greater Value Yet to Be Realized: AI in Pharma Is Far from 'Overcrowded,' Says Dr. Tao Du of Eglen Pharmaceuticals

In the past few years, with the rise of AI technology in the innovative pharmaceuticals field, AI entrepreneurs from both within and outside China have rushed into the sector. Startups have sprung up like mushrooms after rain, while traditional pharmaceutical companies and internet giants have also joined the fray, causing "AI-powered drug discovery" to experience emergence, saturation, fierce competition, and disappointment in a short period.

This fire has continued burning into this year. From the total financing amount of China's AI pharmaceuticals industry surpassing 8 billion yuan in 2021, to briefly experiencing a capital winter in 2022, the emergence of ChatGPT in 2023 has once again placed this field at the forefront.

Nowadays, the value that AI pharmaceutical technology can bring to the research and development of innovative drugs is no longer questioned. However, so far, we have yet to see an AI-developed drug reach the market. On one hand, an increasing amount of capital and entrepreneurs are rushing into the field: If there’s any way to escape the homogenization “involution” and the pressure of the “decade-decade rule” in innovative drug development, AI — with its ability to process and analyze complex data — will undoubtedly play a significant role. On the other hand, breakthrough solutions for computational power, algorithms, and databases in AI-driven pharmaceutical R&D remain unresolved. The extent to which AI can currently enhance the efficiency of new drug development still needs time to prove.

Recently, the first Open AI Developer Conference was held, and AI seems to have become "capable of doing anything." However, in the pharmaceutical field, many questions remain about how AI technology should be applied. To what extent and scope is AI truly useful? For a pharmaceutical industry operating under strict regulatory environments, how should it face industrial development alongside constantly evolving regulatory policies? Amidst the ongoing hype, could an "industry reshuffle" occur? ... With these questions in mind,VCBeat had the privilege of speaking with Dr. Tao Du, former senior reviewer at the FDA and chairman of Shenzhen Evergreen Therapeutics.

Dr. Tao Du, former Senior Reviewer of the FDA and Chairman of Shenzhen Evergreen Therapeutics

VCBeat: To what extent can AI/ML play a role in the pharmaceutical field?

Dr. Tao Du: Artificial Intelligence (AI) and Machine Learning (ML) have been in the pharmaceutical industry for more than 10 years. In the past five to six years, despite differing views on AI and ML, there is consensus that they will become an indispensable technology in the pharmaceutical industry. The point of contention lies in the extent to which AI/ML will play a role in drug development.

In the past, some people mythologized this revolutionary technology, believing it could solve all the problems of the pharmaceutical industry and generate drugs, but this is obviously unrealistic. The FDA and EMA both have clear definitions for AI/ML, and the current consensus is:AI/ML is just an auxiliary tool that can help the pharmaceutical industry develop.

10 years, 1 billion US dollars, 10% success rate—this was the core understanding people had of new drug development in the past. How to reduce R&D costs, shorten R&D cycles, and improve R&D success rates has become an urgent priority for the entire pharmaceutical industry. With the maturation of technologies such as AI/ML,It is also an inevitable trend to apply it to the pharmaceutical industry.

As Eli Lilly CEO David Ricks said, AI is the biggest, revolutionary technological advancement he has seen since becoming an adult. Only two other technologies are comparable: the Internet and the iPhone.

In the pharmaceutical field, it has revolutionized the traditional pharmaceutical industry process, from the discovery of molecular targets to pharmacological and toxicological experiments, and then to each phase of clinical trials. However, we also clearly recognize that AI is not a panacea; it cannot solve all problems in the pharmaceutical industry but can only assist in addressing some of them.

VCBeat: What key points should we bear in mind when considering the application of AI technology to empower innovative drug development?

Dr. Tao Du: The prerequisite is to clarify the purpose of using AI technology, which is to develop drugs with urgent clinical needs. The most important purpose of using AI technology is to accelerate development, reduce costs, increase success rates, and enable patients to access drugs with clinical needs earlier. If the use of AI technology cannot produce drugs with clinical needs, or if the time and cost of drug development are the same as traditional technologies, then the use of AI technology would be utterly meaningless.

VCBeat: Is AI/ML technology currently "overheated" in the pharmaceutical industry?

Dr. Tao DuDuring the COVID-19 pandemic, large foreign pharmaceutical companies纷纷 established their own AI teams. To this day, the world's top ten large multinational pharmaceutical companies have AI teams consisting of dozens or even hundreds of members, developing rapidly. However, AI entered China's innovative drug field relatively late, and due to the communication disruptions caused by the pandemic, today China's AI pharmaceutical industry is mainly concentrated in the drug discovery stage, rather than being applied throughout the entire pharmaceutical process like in multinational companies. This is also what makes people feel...One of the reasons why AI/ML technology is "overheated" in the pharmaceutical field —— mostIn ChinaAI pharmaceutical companies are concentrated in the drug discovery stage — but there is not much demand for drug discovery in the market.

The guidance document recently issued by the FDA divides the application of AI/ML in new drug research and development into five parts, including drug discovery, preclinical pharmacology and toxicology, clinical trials, post-marketing safety supervision, and AI for manufacturing. A large portion of the document mentions the application of AI in clinical trials, an area that few companies in China have ventured into. Clinical research is the longest and most expensive phase in the pharmaceutical R&D cycle. Generally speaking, using traditional innovative technologies to advance the entire process, from Phase I to Phase III, takes nearly a decade. However, if AI technology is applied, taking Evergreen Therapeutics' first drug, EG-301, as an example, the clinical process can be shortened by half, making the development of innovative drugs more efficient.

In China, Evergreen Therapeutics is one of the few companies that坚持将AI应用YuClinicalThe company developing the trial,This is thanks to our clinical office in the United States. Even during the COVID-19 pandemic, we were able to quickly follow up on and adopt the progress and data of AI clinical trials abroad.Since 2020, we have been attempting to apply AI in clinical settings to develop our first innovative drug, EG.-301, while expanding the AI team and integrating it with the pharmaceutical and regulatory teams, it took approximately two years to promote EG.-301Received FDA approval and directly entered Phase II clinical trials.

When advancing the second innovative drug EG-101, we had already mastered relatively mature AI clinical development technology, completing the entire process in just half a year. With the continuous accumulation of experience and ongoing refinement of the technology, it is foreseeable that in the future, our use of AI in clinical drug development will become increasingly efficient.

As the COVID-19 pandemic comes to an end, communication between China and other countries is becoming increasingly deep and frequent. AI technology in China will undoubtedly be integrated into the entire process of pharmaceutical innovation and development—from drug discovery to the drug production stage. After all, if AI pharmaceutical companies only focus on the drug discovery phase, once that phase is over and innovative drug development moves into the preclinical and clinical stages, these AI companies are hardly different from ordinary innovative pharmaceutical enterprises.

In the future pharmaceutical industry, the role of AI/ML will inevitably become more significant, regardless of our intentions.Therefore,AI Technology in the Pharmaceutical Industry Is Not Overheated, but Rather StillNot enoughMaturity, or the coverage of China's AI pharmaceutical industryStillNot enoughExtensive.Given the current situation, AI is still in an early stage in the pharmaceutical industry. In the future, through close collaboration between various AI service agencies and traditional pharmaceutical companies, the application of AI in the pharmaceutical R&D process will become increasingly widespread.

VCBeat: What are the difficulties in applying AI to clinical trials?

Dr. Tao Du: The technical difficulty of using AI in clinical trial phases may be much higher than that of using AI in drug discovery phases.

Currently, AI in the field of new drug research mainly focuses on drug discovery and the prediction and inference of the pharmacology and toxicology of novel molecules. In the clinical stage, it plays a role in the design and prediction of Phase I to Phase III clinical trials. According to current technology,AI in Clinical DevelopmentApplicationBaseThere are three on the book.MainPart,SeparatelyYesSelection of Indications,Patient'sScreening，AndDetermination of Clinical Endpoints。

First,The selection of indications is not the selection of diseases, but the selection of specific time periods for the onset of specific diseases.And SeverityFor instance, the early stage of COVID-19 is viral infection, while the late-stage cause of death is cytokine storm. Therefore, the drugs for treating early and late stages of COVID-19 are definitely not the same. In other words, the early and late stages of the disease are actually different indications, and their treatment methods also differ.

Therefore, in the selection of patients, it is necessary to screen the stage of the disease process they are in, and the confirmation of clinical endpoints must be determined through various examination indicators to confirm the efficacy and safety of the drug.

In the traditional innovative drug development process, clinical phenotypes are usually used to select and confirm indications, patients, and clinical endpoints. After AI intervention, a new reference factor—genotype—will be added to the entire clinical trial design. High-quality data is collected and clinical trials are designed based on "clinical phenotype + genotype." Furthermore, the integration of AI also promotes the development of "reverse translational medicine."

The traditional R&D pathway is the process of putting basic research findings into practice, typically starting with understanding the pathogenesis of a disease, identifying the corresponding target, developing targeted drugs based on the characteristics of the target, and finally applying them to patients. This process is extremely time-consuming and costly. And"Reverse Transformation Science" takes a different approach, starting directly from patients through genomics and clinical phenotype analysis,Using data as the basis for clinical research andFoundation, thereby developingMoreEffective therapy.By contrast, the latter has higher R&D efficiency, but requires developers to have extremely rich clinical experience, medical knowledge, and regulatory background, based on AI data.

Compared to the overall R&D cost of an innovative drug, spending in the new drug discovery phase is relatively low and takes less time. However, clinical development of a new drug represents the longest, most expensive, and least successful part of the entire drug R&D process, with a high risk of failure. Therefore, countries around the world are exploring how AI can be applied in clinical settings, hoping that AI can provide more technical support for clinical development to shorten its duration, increase success rates, and reduce costs.

VCBeat: Are we at the point where we can expect the first AI-developed drug to hit the market?

Dr. Tao Du：Regarding the issue of the "first AI drug," we must first clarify two concepts:"AI+" or "+AI""AI+" is AI-centric, led by scientists to create business value, with a typical example being self-driving cars. On the other hand, "+AI" is based on traditional industries, where AI empowers these industries to generate new commercial value.

For the pharmaceutical industry, AI is an aid, an empowerment, and a reinforcement, but it cannot replace existing pharmaceutical technologies.In the FDA and EMA's recent guidance documents, AI technology has been clearly defined as a tool that can assist, rather than replace, existing drug development processes.

In my view, even if AI can "design from scratch" a new molecule, it will still need to return to the development process of the pharmaceutical industry. The possibility of solely relying on AI to design an innovative drug and skipping clinical trials is slim. It must be based on past knowledge, clinical data, and industry principles. Therefore, strictly speaking,AI Drugs Are Medicines Produced by AI-Assisted Existing Pharmaceutical Technologies.

In the field of new drug development, ClinicalResearchIs aEspecially withResults-orientedThe stage of development. The question that remains for us now isHow to Shorten the R&D Cycle and Reduce R&D Costs, and thenImprove R&D EfficiencyAndSuccess Rate。According to statistics, in 2022, the FDA received over 180 IND submissions for the use of AI in clinical design. I believe this number will increase in 2023. However, it will take some time for clinical trials to show results, and the extent of AI's value in clinical settings remains unpredictable for now. What is certain is that AI will undoubtedly play a critical role in innovative drug development. Within the next 3 to 5 years, we will definitely see case reports on how AI technology improves the success rate of clinical trials.

VCBeat: Are AI pharmaceutical companies "involuting"?

Dr. Tao Du: This superficial "involution" actually reflects a situation of mutual imitation among companies in the AI pharmaceuticals field, where often multiple companies in the same track are doing very similar things. There are approximately 60 to 70 pharmaceutical companies in China targeting the new drug molecule discovery stage. As far as I know, the existence of a large number of AI molecular design service companies will inevitably lower the prices of such companies' service offerings and reduce their commercial profits.

"Because there are still very few companies in China using AI in the clinical stage, for Evergreen Therapeutics, I haven't felt the 'involution' yet."To "advance" to the clinical stage, the team must possess three key capabilities: medical expertise, regulatory knowledge, and AI proficiency.If one of the three key capabilities is missing, AI cannot be effectively utilized in clinical settings at the very least.

In addition, humans have approximately 20,000 to 30,000 diseases, but currently, no more than 1,000 diseases can be effectively treated. Therefore, in fact, there are still many diseases waiting for people to develop new therapies. This is also the reason why AI in the clinical field is far from being "overcrowded" — even though we are all using AI to design clinical trials, each company focuses on different disease treatment areas. This is a very large battlefield, and there are more opportunities in every niche area.

VCBeat: In the face of a possible "industry reshuffle," how can companies clarify their own advantages?

Dr. Tao Du：As a pharmaceutical professional, you must not only focus on China but also look at the entire world.. Only by seeing the whole world can we see more opportunities.

By the end of 2022, when we talked about using AI in clinical settings, many people were still skeptical. But this year, no one questions the role of AI in clinical practice anymore. As we have always emphasized, AI has applications across all fields and stages of the pharmaceutical industry; it’s just that it hasn’t been fully utilized, leading to a concentration in a single stage, which seems a bit crowded. If we broaden our perspective to the entire world and all fields, the industrial capabilities of AI will be more fully revealed.

VCBeat:What is the best business model?

Dr. Tao Du: This is a question of business thinking: value can only be realized by identifying demand. Even if there is demand, entering a market that is too crowded offers no value. At the outset of any project's initiation, Evergreen Therapeutics conducts a commercial evaluation, primarily considering three questions:

First, whether the market size of this drug is large enough, referring to the potential expenditure scale that the market may pay for this drug in the future;

Secondly, whether this drug has many competitive products; the more competitive products there are, the greater the difficulty of competition.

Thirdly, is the development difficulty of this drug too great? Even if the market size is sufficient and there are few competitive products, a product with excessively high development difficulty is not something an ordinary Biotech company can handle.

If we merely copy others' products and business models and rush into a field in a "following the trend" manner, it will inevitably lead to vicious competition, which is not beneficial to the industry or to our own development. Our current leading position in AI clinical applications in China is precisely due to our adherence to a differentiated development approach from the very beginning.