Hong Kong- and New York-based Insilico Medicine on Tuesday announced a drug for treating idiopathic pulmonary fibrosis (IPF) designed by generative artificial intelligence (A.I.) has advanced to Phase 2 clinical trials, which means the drug has been administered to its first human patient.
IPF is a chronic lung disease that makes it more difficult to breathe, starving the body of much-needed oxygen. IPF is currently regarded as incurable, but treatable.
“Generative A.I.” is the level of artificial intelligence that can accept fairly broad commands from a human user and create a complex finished product. Such A.I. systems grow more powerful and useful as they “learn” by accumulating information. DALL-E, the computer art program that can fulfill instructions like “Show me what the Peanuts characters would look like if Picasso drew them” is a popular example.
Creating a new medicine is a daunting task. The design stage includes a great deal of labor-intensive research that could hopefully be completed more quickly by A.I.
Insilico Medicine researchers told China’s state-run Global Times their new INS018-055 drug was designed with such time-saving assistance from generative A.I.:
Compared to a traditional pharmaceutical approach, AI technology greatly accelerates the research and development efficiency. “AI pharmaceuticals can encompass all possible diseases and targets based on a keyword, then proceed with individual identification and diagnosis, and finally select the most valuable parts,” Xu Zuojun, the leading principal investigator of the INS018_055 and chief physician of the department of respiratory and critical care medicine at Peking Union Medical College Hospital, explained.
Pharma A.I. has been looking for a big win, as an initial surge of investor interest faded over the past two years due to a lack of viable computer-designed drugs:
In the future, AI pharmaceuticals represent vast market space. According to BCC Research, the market size for AI-driven drug development is projected to reach $3.117 billion by 2024, with a compound annual growth rate of 40.7 percent.
However, some believed there exists a bubble within the AI pharma industry. The market value of some US AI pharma companies has tumbled by over 90 percent to less than $100 million since 2022, despite 77 AI pharma financing events in 2020. To date, no AI-developed drug has been successfully launched globally.
Insilico Medicine’s founder, Latvian-born scientist Alex Zhavoronkov, told the Financial Times (FT) on Wednesday that his company has reached “the moment of truth” with the first human trial of an “A.I.-discovered and A.I.-designed” drug.
Zhavoronkov said the Phase 2 trial also represents “a true test for A.I., and the entire industry should be watching.”
“Our company, and it’s a big, bold claim, can double the productivity of pretty much every big pharma company,” he said, adding that A.I. is also helpful for finding ideal patients to undergo clinical trials.
Zhavoronkov elaborated that his company’s A.I. systems might be able to cut research and discovery time for new drugs in half, which would knock years of time – and a cool billion dollars off the cost – of developing each therapy. He said his technology has already been licensed to several big pharma companies, including Johnson & Johnson, but Insilico is managing its own flagship drug program and early clinical trials so it can fine-tune the evolving A.I. system.
“It’s extremely important to be able to generate huge amounts of data to ensure that your AI is reinforced but also to ensure that it’s constantly learning,” he said.
FT noted that some analysts believe the current state of A.I. drug research is overvalued and overhyped, although most agree the technology has promise and most pharmaceutical companies have forged partnerships with A.I. firms.
A Harvard Law review of A.I. drug research in March listed several ways in which generative artificial intelligence can be helpful, including swiftly executing countless simulations to select targets for treatment, creating three-dimensional blueprints of drug molecules that could address those targets, simulating tests at speeds far beyond real-time to predict the safety and effectiveness of new medicines, and improving the recipes for drug molecules so they are easier to manufacture.
The Harvard Law paper found third-party investment in A.I. drug research surging to over $5.2 billion in 2021, with over a dozen top firms raising hundreds of millions in funding apiece – but, as the Global Times and Financial Times pointed out, the lack of successful drug product development over the past two years is curbing the enthusiasm of investors.
Insilico’s breakthrough – a drug designed by A.I. to address a target chosen by A.I. – could be what investors need to see in order to get big money rolling again. At the far end of the nascent A.I. drug industry lies a promised land of immensely powerful computers that literally sit around all day dreaming up new drugs, simulating years of tests in a matter of days, and learning from their mistakes faster than an entire generation of human researchers.
Zhavoronkov told Nature that as recently as five years ago, he had trouble explaining to Big Pharma investors exactly what his technology does. He said he does not have to explain its benefits anymore – he just has to prove he can deliver them.
“We’ve managed to get here in three years, and we didn’t fail. And we did it multiple times,” he said of Insilico’s Phase 1 trials. “We have 31 therapeutic programs. In 2020, we had zero.”
Perhaps the most wondrous – and fearsome – detail of A.I. drug technology was Zhavoronkov’s revelation that pharmaceutical scientists at major corporations immediately realized his molecules were designed by neither human hands nor nature.
“They said to me: ‘Alex, these molecules look weird. Tell us how you did it.’ We did something in chemistry that humans could not do,” he beamed, proud godfather of an artificial mind that created molecules that do not exist anywhere else in the universe – and are currently floating inside somebody’s lungs.