Predicting all the biomolecules on Earth, Google AlphaFold 3 is about to disrupt medicine? How big is the potential of tech giants competing for AI pharmaceuticals?

National Business Daily

2024-05-11 15:47Posted on the official account of Sichuan Daily Economic News

Every reporter: Wen Qiao Every editor: Lan Suying

The key challenge in drug discovery, the problem of predicting the three-dimensional structure of proteins based on amino acid sequences, has once again been solved by artificial intelligence (AI)!

On May 8, local time, Google DeepMind and sister company Isomorphic Labs jointly launched AlphaFold 3, which immediately appeared on the front page of Nature magazine. AlphaFold 3 is the third-generation version of AlphaFold, which DeepMind claims can predict the complex structure of almost all molecular types in the "protein database" with unprecedented precision.

Image source: Screenshot of the Nature website

As AI shows amazing potential in the field of drug research and development, the "Daily Economic News" has noticed that it is far more than just Google as a technology giant that is eyeing the AI drug research and development track, and companies such as Microsoft, Amazon, and Nvidia are carrying out or investing in related projects.

With that comes a lot of "hot money". Data shows that since 2021, there have been 281 venture capital deals for global AI drug development startups, with an investment of $7.7 billion. According to foreign media forecasts, the potential market size of AI drug research and development is close to 50 billion US dollars (about 361.3 billion yuan), and it is expected that by 2025, 30% of new drugs will be developed using AI. 

Technology companies are stepping up AI drug R&D

According to the paper published in Nature, AlphaFold 3 is up to 50% more accurate than the best available traditional methods in benchmarking protein interactions with other molecule types, without the need to input any structural information, making AlphaFold 3 the first AI system to surpass physics-based approaches in biomolecular structure prediction.

Why is AlphaFold 3 a revolutionary system? This is because the interaction of proteins (from enzymes essential for human metabolism to antibodies against infectious diseases) with other molecules is key to drug discovery and development.

It has been reported that AI models can be trained on hundreds of millions of different protein sequences and their underlying structures, thus completely simulating proteins, eliminating the need for expensive molecular dynamics simulation calculations. For example, Eli Lilly has been using AI to search millions of molecules, and in just 5 minutes it takes a whole year to synthesize a molecule in a traditional lab, so it makes sense to test the limits of AI in medicine.

As AI shows amazing potential in the field of drug discovery, it is far more than just Google that is eyeing the AI drug development track, and almost all of the tech giants have shown interest in the AI pharmaceutical field, with Microsoft, Amazon, Salesforce and other companies all working on protein generation projects.

According to reports, the last year alone has witnessed the birth of a number of medical-related AI models and tools: Salesforce launched ProGen, a large AI model for protein generation; Microsoft released EvoDiff, a general-purpose AI framework for protein generation; Amazon has released a protein folding tool for its AWS machine learning platform, SageMaker; Nvidia launched BioNeMo, a generative AI cloud service for drug discovery, and invested in pharmaceutical companies Recursion Pharmaceuticals and Iambic Therapeutics.

In addition to the big companies in Silicon Valley piling up to increase AI drug research and development, the explosion of ChatGPT at the end of 2022 has also pushed many AI startups to enter the field.

Deep Pharma Intelligence, a biotechnology market research agency, released an analysis report on the AI drug discovery industry in 2023, which found that 91 of the more than 800 AI-focused companies were established between 2022 and 2023, and most of these companies are AI startups, after in-depth research on more than 800 AI companies, 1,900 investors and other industry participants.

In addition to this, the report says that there is a growing trend of collaboration between big pharma companies and with AI startups. The reporter noted that in April this year, Moderna and OpenAI announced that the two sides are working together to innovate and promote the potential of AI in the medical and health field.

Amgen has also formed an AI-focused partnership with NVIDIA, and Amgen has begun applying predictive models to pharmaceutical processes. Kimberly Powell, vice president of healthcare at Nvidia, said that this allowed Amgen to shorten antibody design from two years to nine months.

It is reported that it usually takes more than a decade to launch a new drug, with an average cost of at least $2.6 billion and up to $6 billion. The advent of AI offers a potential solution for the pharmaceutical industry, where costs are rising and returns are diminishing. According to a report released last year by BCG, a consulting firm, AI can save at least 25%~50% of the time and cost in the drug discovery step in the preclinical stage.

The Impact of AI on Drug Development Time and Cost Image source: BCG report

The market size reaches 100 billion, but the data is still a "hard injury"

With the popularity of AI medicine, "hot money" has also poured in.

According to Pitchbook, since 2021, there have been 281 venture capital deals for global AI drug discovery startups, with an investment of $7.7 billion. Looking further afield, between 2014~2023, AI-driven pharmaceutical investment soared, reaching a cumulative total of $60.3 billion, which also proves the transformative potential of AI.

In the case of Nvidia, for example, Nventures, the venture capital arm of BioNeMo, the company's AI drug discovery platform, has invested most of its money in drug development projects over the past two years. According to the data, seven of Nventures' 19 investment deals were in AI drug discovery startups.

Kimberly Powell said in an interview with the media that healthcare will be Nvidia's next "multi-billion dollar business". Nvidia CEO Jensen Huang has repeatedly stressed that digital biology will be "the next amazing disruptive technology."

According to a report released by Deep Pharma Intelligence last year, nearly 40% of AI companies focused on early-stage drug development make early-stage drug development the dominant field and one of the most trusted from an investment perspective. There are many reasons for this, such as the fact that this is a data-dependent phase that requires virtual screening, iterative learning, and identification and validation of targets (compounds).

Regarding the role that AI plays in the early pharmaceutical process, Kimberly Powell predicts that generative AI may influence discovery by generating ideas that humans would not naturally think of. "[Probably] a protein that has never been made before in nature, or to some extent a protein with therapeutic properties, or a compound that we haven't synthesized yet and doesn't exist in any database in the world."

The use of AI can also be seen in the back-end of drug development, and according to foreign media analysis, it helps to collect data from thousands of documents in preparation for submission to the US Food and Drug Administration (FDA). Similarly, AI may play a role in technology transfer by making data more accessible.

However, high-quality training data remains a challenge in AI-assisted drug development. Greg Yap, a partner at Menlo Ventures, said, "The biggest hurdle is finding reliable training data. These models are being trained on every piece of public scientific data that describes molecular interactions, but some of the data is not very clean and has errors. ”

Once the scientific community can find ways to improve the quality of training sets, the market for AI-assisted drug discovery is expected to take off. According to foreign media forecasts, the potential size of the market is close to 50 billion US dollars (about 361.3 billion yuan), and it is expected that by 2025, 30% of new drugs will be developed using AI. But at present, AI-based drug development is still in its early stages. The FDA has reportedly approved clinical trials of more than 100 drug candidates developed using AI or machine learning to date, but it could take years to become marketable.

View original image 827K

• 

•