Over 10 years experience of Traceability Solutions
By pharmatrax
Category: Technoloy
No Comments
‘For the first time in human history, our ability to collect data on our biology has outpaced our ability to interpret and act on it’
Brendan Frey’s passion for genomics — the science of analyzing and interpreting our DNA — was ignited in 2002. When a family member was diagnosed with a genetic disorder, there wasn’t enough information for doctors to evaluate the full scope of the problem, let alone fix it. “I thought we should live in a better world,” Frey says. “One in which we can accurately detect and treat genetic diseases.”
A decade ago, the odds of that happening anytime in the foreseeable future were decidedly small. But that’s changing as researchers apply Artificial Intelligence to the field of drug discovery and innovation. Frey’s company, Deep Genomics, is at the forefront of this data-driven approach to medical breakthroughs in Canada.
“Drug development has traditionally been a serendipitous activity, like throwing a stick into a tree and seeing if an apple falls,” says Frey, who is also a professor of engineering and medicine at the University of Toronto. “This worked in the early days, but the low-hanging fruit is gone and this traditional approach is leading to more failures, greater delays and rising costs.”
Developing a new medicine is a long, expensive and often demoralizing process. It can take an average of 10 to 12 years, cost upwards of $2 billion and, after all that, the drug in question may not even make it past clinical trials.
“Only eight per cent of discovered drugs are deemed by regulatory agencies to be worthy of approval,” Frey says. “To turn these numbers around, we need to face the wonderful fact that for the first time in human history, our ability to collect data on our biology has outpaced our ability to interpret and act on it.”
This is where AI comes in — it’s the best technology available for automatically analyzing and applying this mounting treasure trove of data. “Researchers can sketch out roughly what the solution should look like, then use AI to fill in the missing pieces, using large amounts of data and computer systems that can learn from examples.”
Deep Genomics has 20 AI systems, says Frey. “Want to discover a new drug target for a new patient mutation? We have a tool for that. Want to design a drug that will address the problem? We have an AI tool for that. Want to check to see if a drug may lead to a bad side effect? We have an AI tool for that, too.”
At first, there was plenty of skepticism about tools like these. But according to a report from Deep Learning Analytics, a data analytics company in Virginia, AI and R&D start-ups raised more than $156 million in the first quarter of 2018 alone.
“What is now a niche is poised to grow into perhaps the leading subsector in BioPharma in the next two to three years, one that will have the greatest transformational impact on the industry,” the report says.
Russ Greiner, a fellow at the Alberta Machine Intelligence Institute and professor at the University of Alberta, has worked in AI for more than 35 years. In addition to reducing the cost of developing drugs, he says tools enabled by AI “might be useful in helping to treat or even cure patients, and to reduce pain and suffering.”
Canada, he adds,” is clearly a dominant powerhouse” in deep learning.
Last year, for example, Toronto-based Cyclica was named one of the top 20 AI drug development companies in the world. In partnership with companies including Merck, Bayer, Eurofarma and WuXi, it is using AI in conjunction with biophysics, statistics and big data to investigate how drugs might act on multiple targets and disease pathways.
“Traditionally, drugs were designed for one target, reflecting a lock and key model where they’re developed to bind to a single protein,” says Cyclica president and CEO Naheed Kurji. “However, a growing body of research has shown that drugs often have hundreds of off-target interactions (i.e. they don’t just bind to a single protein), leading to unanticipated and unwanted side effects. Our goal is to examine all possible proteins in the body that a drug can bind to.”
Kurji says Cyclica’s research could also help scientists better understand why some drugs work differently for different people, speeding the development of new, more precise medicines. “Our goal is to decrease drug discovery time down to two years. It’s ambitious, but we believe by collaborating with pharma and augmenting the capabilities of scientists with breakthrough technologies, we can achieve it.”
Allan Miranda, head of Johnson & Johnson Innovation JLABS in Toronto, says being part of the current paradigm shift in drug development is both “exciting and humbling.”
“Using AI, we have an opportunity to diagnose early, predict the course of a disease and thus treat the disease early, potentially before symptoms manifest,” he says.
Janssen Research & Development is already using AI to identify biological markers that more precisely detect the effects of treatments. The company is also exploring the use of wearable technologies to gather continuous, real-world data in order to identify factors that predict disease progression or relapse.
“The larger goal is to get a fuller understanding of how patients are doing with respect to their symptoms and response to therapies as they go through their daily lives,” Miranda says. “Eventually, using AI in drug development will improve outcomes for patients, as better therapies reach patients sooner — which is good news in a world where ‘patients are waiting.’”
Kurji is similarly enthusiastic about the potential for AI to make drug development “less of a high-stakes gamble” and lead to better, safer drugs for more indications. “Diseases that currently have no treatment, such as Alzheimer’s, and those where treatment is arduous and unreliable, like cancer or diabetes, may move into the realm of treatable or even curable much sooner than they would under the old model.”
However, Kurji adds that AI is not a silver bullet. “AI is most suitable for systems that have an abundance of data and it’s insufficient at creating novel hypotheses that add to that data — AI can’t ‘think outside the box’ on its own.”
Frey agrees, adding that AI also can’t solve problems that aren’t clearly defined. Still, like Kurji, he is optimistic about what the future might hold and has set similarly ambitious goals. Deep Genomics is currently developing therapies for rare metabolic, ophthalmologic and neurodegenerative disorders, with clinical trials expected to start in 2020.
“Instead of an eight per cent success rate, a 13-year development time and a billion-dollar price tag, we’re aiming for a 50 per cent success rate, a four year development time and a 50-million-dollar price tag,” Frey says. “That would be a game changer.”