Isomorphic Labs Announces $3 Billion Drug Discovery Partnerships with Eli Lilly and Novartis

On January 8, Alphabet's AI pharmaceutical companyIsomorphic Labs Announces Two $3 Billion Drug Discovery Agreements with Eli Lilly and Novartis

Through the collaboration with Eli Lilly and Company, Isomorphic Labs will obtain$45 millionThe down payment, and may be eligible for$17 BillionMilestone-based payments. This collaboration involves the discovery of treatments targeting multiple disease-related proteins and pathways.

The partnership with Novartis, where Novartis will pay$37.5 millionThe prepayment, with a total milestone amount of up to$1.2 billionThe focus of the Novartis collaboration is on discovering small-molecule therapies targeting three undisclosed targets.

Spun off from DeepMind, Isomorphic Labs establishes its first pharmaceutical partnership, marking the beginning of a new chapter for the company.

01 First Collaboration, Beyond Drug Development

Demis Hassabis stated that Isomorphic Labs had previously conducted preliminary negotiations with a series of large pharmaceutical companies, but Eli Lilly and Novartis showed strong cooperation intentions.

The primary goal of these collaborations is clear:Manufacturing drugs, rather than pilot projects or academic collaborations.

Eli Lilly and Company Chief Scientific and Medical Officer Daniel Skovronsky and Novartis Biomedical Research President Fiona Marshall were involved in both deals.

Marshall described the collaboration as "a match made in heaven," noting that Novartis has also partnered with other tech giants, such as Palantir for data storage and Microsoft for generative chemistry.

Novartis stated that the targets they chose are difficult, all of which are proteins with unresolved structures or where chemists cannot find the right type of chemical substance.

Isomorphic will use high-performance computing to virtually screen compounds, thereby finding molecules that can hit these targets. Traditional screening typically covers 1 million compounds, while virtual screening can cover 20 billion compounds.

The second approach that Isomorphic Labs plans to explore is de novo design, which involves designing molecules from scratch to match the binding pockets of proteins.

Marshall, President of Novartis Biomedical Research, said in an interview: "We are confident that Isomorphic Labs can achieve the readily attainable result of reducing the time from target identification to obtaining candidate molecules."

"On average, it takes three to four years using traditional methods. I think it's reasonable to halve that time."

But for Isomorphic, the company also hopes to adjust the AI model through cooperation.

Demis Hassabis said that they need to understand how much fine-tuning their model requires to target a specific drug target.

"I don't know what the *balance is between building a general system and building a specialized system around specific data or a specific program," he said. "Right now, I'm imagining it's 50-50, but we'll know soon. Ideally, I'd like the general ratio to be 80-20, as that's more scalable for our company."

Isomorphic Labs hopes that its partnerships with Novartis and Eli Lilly can answer this question: how much can a general model achieve in testing, and in which areas does it need specialized data for specific drug target inputs and model training.

Isomorphic Labs is expanding the functionality of AlphaFold beyond just predicting static protein structures.

"The protein structure itself 'is not very helpful for drug design,' and predicting how small molecules bind to specific proteins is more valuable."

A holy grail in this field is to link the structure of these molecules to their function, and the technology represents the "next phase."

Isomorphic Labs is studying other models, which are still confidential, that could help predict certain properties of compounds, such as pharmacokinetics and potential toxicity.

In addition to these collaborations, Hassabis stated that the goal for 2024 is for Isomorphic’s machine learning team to continuously improve the accuracy of its predictive systems while the company also advances its internal pipeline.

Details about its own drug programs remain confidential, but Isomorphic has narrowed down a list of about 20 targets to a few, and the company is utilizing CROs to manufacture and test potential compounds.

"We are still in the early stages of testing the platform's predictions," said Demis Hassabis. "How reliable are the AI models? Once we believe they are mature, then we can really push forward with internal projects."

02 Originated from DeepMind

In the 2020 CASP14, AlphaFold2 stood out by predicting with atomic accuracy. Even for proteins lacking templates, it can produce outstanding results within minutes. This is a method capable of constructing high-resolution predictions without knowing similar structures.

In July 2021, DeepMind publishedPredictions of the structures of almost all human proteins. In just two years, the influence of this paper has almost surpassed 100,000 research papers published by *Nature* since 1900, ranking 50th, and has been cited by more than 7,000 papers in journals.

According to the significant achievements of AlphaFold, on November 4, 2021, Alphabet, the parent company of Google, established a new drug research and development company named Isomorphic Labs in the UK, with Demis Hassabis serving as the Chief Executive Officer.

The mission of Isomorphic Labs is to "reimagine the entire drug discovery process from scratch" and ultimately find treatments for some of humanity's most devastating diseases.

In the first year or two after its establishment, aside from the executive information disclosed on the official website, Isomorphic Labs released very little news.

Starting in the second half of 2023, Isomorphic Labs began to make significant progress in both technology and business.

On September 19, 2023, the cover article of *Science* magazine featured AlphaMissense, a pathogenicity prediction model for missense variants derived and improved by the DeepMind team based on AlphaFold.

AlphaMissense analyzed a total of 7,100 possible missense variants, of which 89% were classified, with 32% potentially pathogenic and 57% likely benign.

Two days later, on September 21, it was hailed as a Nobel Prize barometer"2023 Lasker Award", awarded to Dr. Demis Hassabis, CEO of DeepMind, and Dr. John Jumper, representing recognition from the mainstream scientific community.

In October 2023, Google Deepmind, in collaboration with Isomorphic Labs, released a new generation of the AlphaFold model, which is expected to change the game rules for drug discovery.

The new generation of AlphaFold can predict the structure of virtually any molecule in the Protein Data Bank (PDB) with atomic precision, including ligands (small molecules), proteins, nucleic acids (DNA and RNA), and biomolecules containing post-translational modifications (PTMs).

The model’s extended capabilities and performance can help accelerate breakthroughs in biomedicine and usher in the next era of “digital biology,” providing new insights into disease pathways, genomics, potential therapeutic targets, drug design mechanisms, and the functionality of new platforms for protein engineering and synthetic biology.

According to the UK's Financial Times, Isomorphic Labs was accelerating its operations by poaching pharmaceutical talents and opening new offices at the end of last year.

Team 03 now has 90 members

The team lineup of Isomorphic is very impressive. The company currently has approximately90 employees。

Executives include Chief Scientific Officer Miles Congreve, and several employees have also joined from companies such as DeepMind, BenevolentAI, Google, and AstraZeneca.

Demis Hassabis, CEO, graduated from University College London, founder of the startup DeepMind, British game developer, neuroscientist, and artificial intelligence entrepreneur, known as the "Father of AlphaGo."

Chief Scientific Officer Miles Congreve, who has helped design 20 drugs that have undergone clinical evaluation and is a co-inventor of Kisqali® (Ribociclib), an approved breast cancer treatment.

Chief Artificial Intelligence Officer Max Jaderberg, leading the Open-Ended Learning Research Team, has pioneered numerous algorithms combining large-scale deep learning, reinforcement learning, and generative models to achieve world-class results in artificial intelligence. Prior to this, he served as the CEO and co-founder of Vision Factory.

Chief Technology Officer Sergei Yakneen, with over 20 years of experience in engineering, machine learning, product, life sciences, and medical research.

Scientific Advisory Board member Dr. Jennifer Doudna, winner of the 2020 Nobel Prize in Chemistry and founder of Caribou Biosciences, studied under Thomas Robert Cech, recipient of the 1989 Nobel Prize in Chemistry.

Scientific Advisory Board member Dr. David MacMillan was awarded the 2021 Nobel Prize in Chemistry for developing asymmetric organocatalysis, a new tool for precise molecular construction.

Scientific Advisory Board member Dr. Paul Nurse was awarded the 2001 Nobel Prize in Physiology or Medicine for his discovery of protein molecules that control cell division.

Scientific Advisory Board member Dr. Venki Ramakrishnan was awarded the 2012 Nobel Prize in Chemistry for his research on the structure and function of ribosomes.

The company said that Isomorphic will hire more employees next year.

A Fire in the Cold Winter

AlphaFold Ignites the Entire AI Pharmaceutical Track: Capital Flows In, Startups Emerge Endlessly, and Various Stories Unfold.

The birth of Isomorphic also represents the implementation from AI models to the pharmaceutical industry.

However, in the past two years, the industry's bubble has gradually burst, the story of cost reduction and efficiency enhancement has started to be questioned, and drugs in the clinical stage are also facing the challenge of insufficient efficacy.

CEO Demis Hassabis once said, "If we are right and can achieve another six AlphaFold-level breakthroughs, we could reduce the time required for drug discovery by an order of magnitude, perhaps both cost and time, and achieve a higher success rate in the next phase."

Today, Isomorphic Labs, as an AI pharmaceutical company that has gathered the industry's top talents, hopes to lead the way out of the industry’s technical and commercial challenges.

Let us remain rational and hopeful.

[This article is authorized for release by Zero2IPO's WeChat Official Account: PharmaSmart. This platform only provides information storage services.] If you have any questions, please contact (editor@zero2ipo.com.cn) for handling by Zero2IPO.