"It takes more than 10 years and over $2 billion (2 trillion won) in expense to develop a single new drug and get it approved. The failure rate tops 90%. The moment a new drug candidate discovered by AI receives its first approval, the pharmaceutical industry will shift to a sustainable structure."
Mati Gill, CEO of AION Labs in Israel, said in an interview with ChosunBiz at Walkerhill Hotel in Seoul on the 28th, "The true value of AI drug development has yet to be proven, but within the next five years the first AI-approved new drug will emerge." Gill visited Korea to attend DIPS Global TechCon, the official business event of the Asia-Pacific Economic Cooperation (APEC).
AION Labs is an AI-based new drug development consortium jointly established by global big pharma including U.S. Pfizer, U.K. AstraZeneca, Germany's Merck, and Israel's Teva, along with Amazon Web Services (AWS), the world's largest cloud (virtual server) service corporations. Gill previously served as chief operating officer (COO) of the legal institutional sector and director in charge of government affairs at Teva. Gill has experience across Israel, the Asia-Pacific, Europe, the Middle East, and South America.
Gill said, "AION Labs differs from traditional venture capital (VC) or accelerators (AC), which launch startups and recover investment and revenue through mergers and acquisitions (M&A) or initial public offerings (IPO)," adding, "We select common challenges in new drug development and gather researchers to directly establish and nurture startups in a 'venture studio' model."
Unlike VCs, a venture studio is a corporations model that actively participates across the entire company-building process, from ideation to funding, management support, and talent acquisition. AION Labs solicits new drug development ideas from researchers, selects technologies, and then launches and grows startups based on those technologies.
So far, nine startups have been created through this approach. Initial funding per startup is about $1 million (1.43 billion won). Startups founded by AION Labs receive regular monthly advice and consultation from institutional sector experts at big pharma such as AstraZeneca, Pfizer, Merck, and Teva.
In conventional new drug development, just discovering candidates in the initial stage takes years and hundreds of billions of won, yet the success rate is below 10%. When AI is combined with new drug development, it can learn vast biological and chemical data to quickly identify promising candidates, reducing both the development period and expense while boosting the chances of success.
Israel took note of AI's potential and moved first. In 2018, the Israeli government designated the "bioconvergence" industry, which combines AI, Machine Learning, and computing technologies, as a next-generation national strategic industry and pursued the establishment of innovation labs through public-private cooperation. Although several pharmaceutical companies expressed interest at the time, considering the complexity and technical difficulty of new drug development, they chose a collaborative model rather than competition.
As a result, a public-private partnership was formed around the Israel Innovation Authority, bringing together four big pharma companies, AWS, and venture capital (VC). This partnership officially launched AION Labs in 2021. The goal is to combine the participants' technology, capital, and data to solve the structural limitations of new drug development.
Gill said, "Teva adopted advanced technologies such as AI, Machine Learning, computational biology, and computational chemistry to find new growth engines after its flagship multiple sclerosis treatment Copaxone, but its internal capabilities were lacking," adding, "To overcome this limitation, we founded AION Labs as an innovation model that collaborates with global pharmaceutical companies."
AION Labs operates startups on two tracks: "problem-centric" and "technology-centric." A representative example of the problem-centric track is DenovAI. The company began with the question, "Can AI design new antibodies that do not exist in nature?" The core is "denovo" technology, which creates entirely new proteins from scratch without referencing existing proteins.
Gill said, "About 90% of diseases cannot be treated with existing antibodies," adding, "When we presented this challenge to researchers worldwide, technology from the European Molecular Biology Laboratory (EMBL) was selected, and DenovAI was founded around its lead scientist, Dr. Kashif Sadik." DenovAI aims to enter its first clinical trial within three years.
A representative example of the technology-centric track is Cassidy Bio. It uses a large language model (LLM) to design guide RNA for CRISPR gene editing. Guide RNA is a key factor that determines accuracy and safety in gene therapy.
A CRISPR gene editor is not an actual pair of scissors but an enzyme complex that cuts the desired gene. When the guide RNA recognizes and binds to the target deoxyribonucleic acid (DNA) segment to be cut, the Cas9 protein binds to the DNA and cuts it. A large language model is an AI technology, like ChatGPT, that learns massive amounts of textual data to understand language and generate user-requested content. Cassidy Bio aims to design optimal guide RNA with AI.
Recently, it has also been collaborating with Korean corporations. Since Aug., it has been conducting an AI new drug development project with Dongbang FTL, a domestic active pharmaceutical ingredient manufacturing corporations. When Dongbang FTL provides protein candidates that could serve as drug targets, Prophet, an AION Labs startup, uses AI to select small-molecule that can bind. The project has been filed as the first pilot initiative of the Korea-Israel joint fund.
Gill said, "We can partner with corporations that have antibody technologies like Dongbang FTL but need AI capabilities," adding, "During this visit, we have meetings scheduled with Korean corporations and investors, and we plan to expand cooperation."
Currently, about 60 AI-designed new drug candidates are in clinical trials worldwide, but none has yet been approved by the U.S. Food and Drug Administration (FDA). After an initial AI drug development boom passed, investor enthusiasm has also cooled.
Gill said, "Now it is time to move toward a sustainable growth phase," forecasting, "If just one or two AI drugs win FDA approval, the industry landscape will change completely." Gill added, "The FDA also spurred change last year by notifying companies to prove their 'AI-based antibody discovery methods,'" and said, "Once the first approval case emerges, the world will refocus on the field, and investment and development will follow like a flood."