Canadians aren’t worried enough about superbugs. As bacteria develop immunity to the antibiotics in our limited arsenal, new treatments will be needed—and soon. Otherwise, routine medical procedures will become impossible, common infections will turn fatal and tens of millions of people will die every year of once curable infections. Superbugs are teeny-tiny time bombs. Thankfully, researchers like Jon Stokes are racing to get new antibiotics to the finish line faster.
Stokes, a 34-year-old assistant professor of biochemistry at McMaster University in Hamilton, Ontario, runs two complementary projects. One is Stokes Lab, a laboratory where he and a cohort of undergraduate and graduate students use artificial intelligence to dramatically hasten the pace of antibiotic discovery. The second is his Biochem company, Stoked Bio, which he founded in 2023 to commercialize the lab’s findings—and in so doing, replenish an antibiotic pipeline that’s close to emptying out. Here, Stokes speaks about his work, the high-stakes fight against antibiotic resistance and why medical AI could mark the turning point in our battle against the bugs.
How big of a deal is antibiotic resistance?
Can I blow your mind? Most of the antibiotics we use today were discovered in the ’40s, ’50s and ’60s, from microbes that live in dirt. Starting in the mid-’60s, our ability to discover novel antibiotics that way decreased. The last brand-new antibiotic we discovered was in 1987; since then, antibiotic resistance has grown, because the more we take them, the more that bacteria evolve resistance. That’s a big problem, because antibiotics are the foundation of modern medicine.
Let’s say you get cancer, for example, and you need immunosuppressive chemotherapy. The only reason we can do that safely is because you’ll be administered an antibiotic to prevent infection. Similarly, if you must have invasive surgery, like an organ transplant, you’re given antibiotics. If you have a preterm birth, the baby will be born without a functioning immune system, and it will be given antibiotics, so it doesn’t die. My family is predisposed to arthritis, so there’s a very good chance that, in the next 20 years, I’ll need a new hip or something. When I see my orthopedic surgeon, she might say, “Hey Jon, I know you need a new hip. We have one; it’s sitting on this shelf right here. But if I give it to you, you might die from an infection that I can’t treat. So, I’m going to give you a wheelchair instead.” As soon as we lose antibiotics that can be reliably deployed, all these modern medical procedures stop, because they become too risky.
Is this already happening in hospitals and clinics?
Yes, and it’s going to get worse. There are roughly 100 different antibiotics available for clinical use, and we already see patients in clinical settings with infections that are resistant to every single one of them. We cannot treat those patients. We can make them comfortable, but we can’t treat the actual problem. This year, 1.5 million people worldwide will die from a drug-resistant infection that, not long ago, we could have cured. More people will die, life expectancy will drop, infant mortality will rise, and quality of life will go down.
Why haven’t we been able to find new antibiotics in so long?
Scientists have tried. We’ve fiddled with the structure of pre-existing antibiotics to eke out a little more from them. There’s a method called high throughput screening, where we take giant libraries of synthetic chemicals—millions of them—and systematically test them to see if they’ll kill the bacterium we want. We tried that for 30 years; it didn’t work. We exist today in what people in my field call a “discovery void.”
You’re working to fill that void with AI and machine learning. Mind explaining how?
Identifying a molecule that can kill a bacterium is simple. But developing a molecule that will kill a bacterium, which is safe, gets to the right spot in your body and can be safely eliminated from it? That’s really challenging. With machine learning, we can run through millions—or billions—of chemical structures that have the potential to kill those bacteria, in a fraction of the time it would take in the laboratory. Just a couple of weeks ago in the lab, we curated a set of 12 million molecules and ran them through one of our AI models. It took about four days. If we had to individually test every chemical in the lab, it would take us 80 years and cost hundreds of millions of dollars. Using our AI models, it cost us probably $10 in electricity. That efficiency is important, because it means we can discover new antibiotics, and the process will be dirt cheap.
Why is the cost so crucial?
Traditionally, an antibiotic cost about $1 billion to develop, give or take. Pretend you’re an investor. You have a billion dollars burning a hole in your pocket. You come to me, and you say, “Hey Jon, you what have you got for me?” And I say: we have the best antibiotic you’ve ever seen. But it’s going to be used for only the sickest patients in the world, for whom every other antibiotic has failed, and they’ll only be on it for a week. This investor could spend a billion dollars to develop this drug, which will be used infrequently, and for a short period of time. Or they could put their money into a blood-pressure medication millions of people will use every day for 30 years. It’s obvious which one they’ll invest in. As well, large pharmaceutical companies aren’t making significant investments in antibiotics because there’s no way to recoup them. Antibiotic development is not an investment; it’s a donation for a public good. And there hasn’t been enough social or economic pressure to encourage more of it.
The questions I find myself asking are, “How bad does this have to get? How many people need to die? And how much does it have to cost us?” We’re already blowing $1 billion a month, just in Canada, due to antibiotic-resistant infections. That includes hospital treatments, lost productivity and all the associated costs on our health system. If a new antibiotic costs a billion dollars, then for the price of lost GDP over two years, we could have more than 20 new classes of antibiotics. The total GDP of the G7 countries is about $50 trillion; $20 billion is less than one per cent of that. The seven richest countries in the world can surely scrounge together the money to fix a global problem for the next century.
There are several drugs in the pipeline at your lab. What are you working on now?
Several antibiotics, as well as a brain-cancer drug that was predicted by one of our AI models to selectively kill glioblastoma, a bad type of brain cancer. We’re trying to develop it right now in collaboration with Sheila Singh, a world-renowned brain surgeon and cancer researcher at McMaster. I’m also quite interested in narrow-spectrum antibiotics. Most current antibiotics are broad spectrum, meaning they kill every bacterium they encounter. When patients take them, they also wipe out the good bacteria in their gut, which opens them up to secondary infections and a bunch of other health issues. Narrow-spectrum drugs spare a lot of this microbiome, so you only kill the pathogen you’re interested in.
What else makes your lab so unique?
Rather than throw my invention over to an external company and hope they develop it, I’m throwing the invention over to my own company, Stoked Bio. That gives us a lot of intellectual continuity between the invention side and the innovation side. Stoked Bio is focused a lot on de-risking. That means developing discoveries so they’re attractive to a large pharmaceutical partner to drive them through clinical trials, which are very expensive to run. So, a big pharma company can look at a chemical we’ve invented in the lab—and that we’ve developed at Stoked—and see that in early tests it’s working how it’s supposed to. Then that chemical is considered de-risked. A large multinational pharmaceutical partner is more likely to work with us at that point. Drug development is a huge effort, and often necessitates close collaborations.
With AI, can that de-risking process be quicker as well?
Yes, we can work through those data sets faster. These algorithms enable us to more rationally and more confidently explore huge regions of chemical space. I’m talking billions, hundreds of billions, trillions of chemicals—more than the total number of atoms in our solar system. It also allows for more collaboration. There’s a company in Utah called Curza, and all they do is antibiotics. They’ve partnered with Stoked to access our computational expertise. In the future, they’ll be able to build off some of that code without starting from scratch. That would make me so happy.
How do humans’ factor into the equation? Are we needed at all anymore?
Absolutely. The whole idea of using AI is to go through tons and tons of BS and allow us to float up to the top the relatively small number of molecules most likely to have all the properties we need. That process—finding a new chemical that kills a bacterium, then figuring out what it does to make that bacteria die—can take a year to two years of sustained work. Now we have an AI model that can predict that biological function in 100 seconds—but it still takes two of my Ph.D. students eight months to validate those predictions. That’s where the people come in. AI streamlines the process, but we still need human beings to take it the rest of the way.
Bio:
Dr. Bill Pomfret of Safety Projects International Inc who has a training platform, said, “It’s important to clarify that deskless workers aren’t after any old training. Summoning teams to a white-walled room to digest endless slides no longer cuts it. Mobile learning is quickly becoming the most accessible way to get training out to those in the field or working remotely. For training to be a successful retention and recruitment tool, it needs to be an experience learner will enjoy and be in sync with today’s digital habits.