Biossil Inc. co-founders Alexander Mosa, left, and Anthony Mouchantaf at their office in Toronto, on March 30.Cole Burston/The Globe and Mail
For years, several well-funded biotechnology companies have used artificial intelligence to create molecules they hoped could become billion-dollar pharmaceutical drugs. None have yet made it to market.
A pair of Toronto entrepreneurs are taking a different approach to combining AI and medicine. Instead of conjuring up molecules, they’re using algorithms to scavenge through the expanse of drug candidates that failed human trials at advanced stages. If their technology can detect ways to give those molecules a second chance, they hope to repurpose them into successful therapies, foregoing years of earlier-stage studies and hundreds of millions of dollars in costs.
That’s the idea behind Biossil Inc., founded in 2023 by Anthony Mouchantaf, a lawyer-turned tech entrepreneur who previously headed Royal Bank of Canada’s venture capital investment strategy, and Dr. Alexander Mosa, who trained to be an internal medicine specialist before leaving to start the company.
They’ve built a software platform using OpenAI’s large language models to uncover promising molecules from the pharmaceutical discard pile, and set out to buy or license them from their owners. They’ve operated surreptitiously in what startup founders call “stealth mode.”
That is, until now.
How the promise of AI is taking hold at Canada’s biggest banks
In the past three years, Biossil has bought or licensed 10 molecules from their owners, Mr. Mouchantaf told The Globe and Mail in an exclusive interview. Two are in advanced clinical trials, and Biossil hopes to get conditional approval from regulators to take three resuscitated drugs to market.
Biossil has partnered with more than 20 universities and research hospitals in Canada, the U.S. and Europe including Toronto’s Hospital for Sick Children, Harvard University, the Mayo Clinic and Denmark’s Aarhus University.
“We’ve very quietly become the most advanced drug developer of this AI era, bar none” said Mr. Mouchantaf, the company’s chief executive officer.
And Biossil has raised substantial financing from venture capital luminaries drawn to its ambition to become Canada’s first homegrown pharma giant.
“We saw a massive vision to disrupt the way drugs are brought to market,” said Janet Bannister, whose Toronto-based Staircase Ventures led a $3.7-million seed financing of Biossil in 2023, backed by fellow Canadian financiers Golden Ventures and Panache Ventures.
Silicon Valley billionaire Peter Thiel’s Founders Fund – an early backer of Vancouver’s AbCellera Biologics Inc. – led Biossil’s US$22-million financing in 2024, then co-led a US$43-million financing last fall, alongside OpenAI. Biossil is now valued at more than US$100-million.
“The thesis is highly ambitious but grounded in a practical understanding of how value is actually created in drug development,” said Founders Fund partner Amin Mirzadegan in an e-mail.
Opinion: Are we approaching the ‘Silent Spring’ of artificial intelligence?
Ian Hathaway, a partner with OpenAI’s Startup Fund, said, “Biossil’s approach to realizing this vision stood out to us immediately, not only for its creativity and ambition, but for its credible path to delivering new therapies,” relatively quickly.
Using AI to design molecules may be groundbreaking technologically but it does nothing to shorten the decade or so of development work required to satisfy regulators that a novel drug is safe and effective.
Mr. Mouchantaf said he and Dr. Mosa “felt any application of technology needed to be married with some fundamental new insight” that innovated on the laborious, costly and often unsuccessful process of bringing drugs to market.
Dr. Mosa was also aware that many drug candidates fail for two reasons: Some only work as intended in certain patients with particular attributes, and trials are sometimes poorly designed or pursue the wrong outcomes.
“It seemed there was an opportunity to mine this reservoir of drugs” that had failed despite passing safety and early efficacy studies, identify the ones with the most commercial promise, and pick up where their previous owners had left off, Dr. Mosa said. Mr. Mounchantaf estimated the previous owners of molecules in Biossil’s portfolio had spent more than US$1-billion developing the drugs before abandoning them.
Mr. Mouchantaf and Mr. Mosa's company Biossil has partnered with more than 20 universities and research hospitals in Canada, the U.S. and Europe.Cole Burston/The Globe and Mail
Other AI drug companies such as Isomorphic Labs, spun off from Google DeepMind’s Alphafold project, use their technology to predict the shape of proteins and where molecules can bind to them to design drug candidates.
Biossil doesn’t concern itself with these structural matters. Instead, it starts with words. The LLMs powering chatbots chop up words and assign each component a string of numbers, or what’s called a vector. By charting the distances between these vectors, AI models can glean definitions and context.
Biossil applies the same idea to medicine. Its AI reads publicly available information about drug candidates, including research, raw data, press releases and securities filings to create written descriptions of each molecule’s attributes and functionality. It then uses something called an embedding model to convert these text descriptions into numbers that can be plotted. Biossil repeats the process for genes that underpin various diseases and plots these numbers against the drug data.
Biossil maps the distances between points in these constellations of data to determine which drugs are the best candidates for addressing traits associated with certain diseases. The company first tested its process on drugs that have already been approved and found that it was an effective prediction method.
The founders say they’re agnostic about what drugs to pursue, but are focusing on treatments for unmet needs for patients with life-threatening or debilitating diseases. They’re keen to produce drugs for underserved populations, figuring that if they can bring new medicines to market less expensively, they can also charge less for them.
Two of their early molecules target sickle cell disease, a debilitating genetic condition that predominantly affects people of African or Indian descent. Johnson & Johnson and Pfizer Inc. each attempted to develop the drugs with biotechnology startups but abandoned those programs after failing late-stage efficacy trials.
2024: Revolution, interrupted: Why AI has failed to live up to the hype in drug development
Biossil’s AI found reasons to revisit both. The drug associated with Johnson & Johnson, called Senicapoc, failed to demonstrate significant improvement over placebos in relieving pain, the study’s main goal. But the data showed it was much better at preventing the breakdown of red blood cells which would cause anemia and other serious complications.
“That tells us the endpoint chosen in retrospect was mistaken,” said Biossil adviser Dr. Isaac Odame, head of hematology and oncology at Toronto’s Hospital for Sick Children and founder of the Global Sickle Cell Disease Network. Biossil obtained the molecule and got Health Canada approval in 2025 to study its impact on red blood cell breakdown in a late-stage human trial.
The goal of the Pfizer drug, an antibody called Rivipansel, meanwhile, was to unblock pain-inducing clogged blood vessels in sickle cell patients. Biossil’s technology revealed the drug worked much better at reducing pain if administered within 24 hours of onset. Because some patients in the failed trial received it later on, it wasn’t nearly as effective for them, which hurt overall results.
Armed with its narrowed data set focused on those who received the experimental drug within 24 hours, Biossil is seeking conditional approval from Health Canada to take the former Pfizer drug to market and has obtained approval from the U.S. Food and Drug Administration to do a confirmatory late-stage human trial.
Dr. Kevin Kuo, a sickle cell expert with Toronto General Hospital, said he was amazed when he learned Biossil’s methods could uncover insights that rivalled what he had learned during his two-plus decades in the field.
“This to me is proof this technique works with other diseases as well,” he said. Last year, Dr. Kuo joined Biossil as its head of medical, the first time he’s worked for a startup. “I’m doing this purely out of conviction.”