Generative AI platforms drive drug discovery dealmaking

Generative artificial intelligence (AI) burst onto the public stage in 2022 with the remarkable text-generating capabilities of the large-language model ChatGPT. Now, the potential of generative AI models to accelerate the design of drug candidates looks set to boost the funding raised in recent years by AI-focused companies and is driving dealmaking by major biopharma companies seeking to enhance their pipelines (Fig. 1 and Table 1, compiled using data provided by DealForma).

A case in point is Xaira Therapeutics, which emerged in April this year with $1 billion of committed funding—one of the largest initial biotech funding rounds ever—and a high-profile management team. The company has the vision of end-to-end application of AI tools in drug research and development (R&D), including a class of generative AI tools called diffusion models to design biologic drugs such as antibodies. Generate:Biomedicines, a company founded by Flagship Pioneering in 2018 that has raised around $750 million since, is also applying such tools in antibody design.

On the dealmaking side, the potential to use generative AI tools to enable the design of antibodies for challenging targets such as membrane proteins or to identify drug candidates with best-in-class characteristics for established targets has been attracting large biopharma companies to partner with AI-focused companies. In November last year, Almirall announced a partnership with Absci potentially worth up to $650 million to use its generative AI platform to identify antibodies for two therapeutic targets in dermatology (Table 1). The following month, Absci agreed another collaboration, this time with AstraZeneca to design an antibody drug candidate for an oncology target. Absci could receive up to $247 million under the terms of the deal.

Absci is also pursuing its own candidates, with its first being an AI-designed antibody against the inflammatory disease target, tumor necrosis factor-like cytokine 1A (TL1A), which is currently in studies to support an investigational new drug (IND) application. The company is hoping to compete with the leaders in the race to bring a TL1A-targeted antibody to market for inflammatory bowel diseases, which has seen three large pharma companies—Merck & Co., Sanofi and Roche—all sign major deals in the past year or so to acquire or license a TL1A inhibitor in clinical development. In May 2023, Merck & Co. paid $10.8 billion to acquire Prometheus Biosciences and its phase-3-ready TL1A-targeted antibody PRA023. Then, in October 2023, Sanofi and Teva announced a deal involving an upfront payment of $500 million to Teva and up to $1 billion in milestones around Teva’s TEV-’574, a TL1A-targeted antibody in phase 2 trials. Later the same month, Roche joined the fray when it agreed to pay $7.1 billion upfront and a near-term milestone payment of $150 million to acquire Telavant and its phase-3-ready TL1A-targeted antibody RVT-3101.

Other AI companies involved in major deals recently include BigHat Biosciences, which entered a partnership with AbbVie with a headline value of $355 million to develop therapeutic antibodies for oncology and neuroscience applications in December 2023 (Table 1). And in a deal potentially worth more than $1 billion between BioMap and Sanofi, announced in October last year, the companies will partner on the development of AI models to enable biologic design and development, such as BioMap’s xTrimo, a large language model for protein structure prediction.

Sanofi also signed a deal with Aqemia in December around the application of generative AI for small-molecule drug discovery. Aqemia, which is eligible to receive up to $140 million in upfront and milestone payments under the agreement, will collaborate with Sanofi using its platform based on generative AI and ‘deep physics’ to identify small-molecule drug candidates in several therapeutic areas (Table 1).

Another company involved in a duo of recent AI-related deals—announced on the same day in January at this year’s JP Morgan conference—is Isomorphic Labs. The company was spun out of Google’s DeepMind subsidiary in 2021 to work on drug discovery applications of AlphaFold, the AI-based protein structure-prediction tool that took the field by storm following its publication in 2021¹. In one deal, Eli Lilly agreed to pay $45 million upfront and up to $1.7 billion in milestones to apply Isomorphic’s technology to discover small-molecule drug candidates for multiple undisclosed targets. And in the other deal, Isomorphic received an upfront payment of $37.5 million and is eligible to receive milestone payments of up to $1.2 billion from Novartis to work on drug candidates for three undisclosed targets.

September saw another same-day announcement of two AI-related collaborations, in this case set up between Merck KGaA and two AI-focused companies, Exscientia and BenevolentAI (Table 1). The deal with Exscientia is focused on applying this company's AI-based drug design capabilities to identify small-molecule drug candidates in the oncology, immunology and neuroinflammation areas, and the terms include a $20 million upfront payment to Exscientia and up to $674 million in milestone payments, as well as potential royalties on resultant products. The partnership with BenevolentAI is aiming to apply their platform to discover drug candidates for three targets in oncology, immunology and neurology. In this case, the potential value of the deal could be up to $594 million, including an undisclosed upfront payment to BenevolentAI and milestone payments. Sales royalties on commercialized products would also be payable.

Insilico Medicine, one of the pioneers in the application of generative AI in small-molecule drug discovery, has also signed a couple of major deals in the past 12 months (Table 1). In September 2023, the company agreed to license an IND-ready small-molecule inhibitor of an anticancer target, the deubiquitinase ubiquitin carboxyl-terminal hydrolase 1 (USP1), to Exelixis for $80 million upfront. Insilico Medicine is also eligible to receive milestone payments and royalties on future sales. This was followed by another deal for a preclinical small-molecule inhibitor for an anticancer target in January 2024, this time the histone acetyltransferase KAT6A, with Stemline Therapeutics. The total potential value of the deal to Insilico Medicine could be more than $500 million, including an upfront payment of $12 million.

Insilico Medicine made the news this year as well with a publication around its leading pipeline candidate INS018_055, a small-molecule inhibitor of the novel antifibrotic target TRAF2- and NCK-interacting kinase (TNIK), which was also identified using the company’s end-to-end platform². The compound is currently in a phase 2 trial for idiopathic pulmonary fibrosis—an indication for which new therapies are greatly needed and that has historically seen many failures in drug development.

While AI-designed antibodies are yet to reach the clinic, the outcomes from trials for small molecules such as INS018_055 that are now in progress will be closely watched to understand whether AI is able to accelerate not just the discovery of drug candidates, but also the identification of those that are capable of succeeding in the clinic where others have not in the past.