With the search for the next medical miracle constantly ongoing, here’s a look into the role of artificial intelligence in drug discovery.
Drug discovery is time-consuming and expensive, but artificial intelligence could save humans both time and money to bring new therapies and cures to patients who need them.
According to PhRMA, the pharmaceutical industry research and advocacy group, the average drug takes over 10 years and $2.6 billion dollars to reach the market. The group also notes that only 12% of drugs entering clinical trials result in a medicine that's approved by the U.S. Food and Drug Administration.
Research and development of pharmaceuticals relies on a lot of trial and error, according to Dr. Yuan Luo, who is Director of the Institute for Artificial Intelligence in Medicine at Northwestern University's Feinberg School of Medicine.
"That’s a bet," said Dr. Luo. "What if it failed? Then years of the investment got wasted."
Dr. Luo and other experts who watch the emerging field of AI in drug discovery said machines will help humans cut down on the time it takes to complete repeat tasks.
"We can move the most promising candidates to clinical trials and save time and save resources," said Dr. Luo.
One Chicago company focused on cutting the time and expense it takes to find a new drug is Evozyne.
Local
"AI makes the process of drug discovery wildly more efficient and much, much faster," CEO Mike Gamson said.
Based in Chicago's Lincoln Park neighborhood, Evozyne creates novel proteins that may lead to new therapies and cures.
Feeling out of the loop? We'll catch you up on the Chicago news you need to know. Sign up for the weekly Chicago Catch-Up newsletter.
"Proteins are nature's machines," said Gamson. "There are these wildly intricate and incredible, tiny machines that do all the work in your body, and really all of the work in the natural world."
Gamson said humans have discovered millions of proteins, but there's many more still undiscovered.
"Imagine that you were zooming around...something as big as a solar system, and you were looking for a grain of sand," said Gamson. "That's about the right mental image to understand the range of possibilities for what a protein could be. As many grains of sand fit in the solar system, that many different kinds of proteins exist."
Over time, humans have learned a lot about the rules of proteins. According to Gamson, scientists can teach those rules to a machine to help humans find new proteins faster.
"AI...allows us to be very focused and use an engineering approach to explore that huge design space, to come up with a solution that we never would have found by random selection," Gamson said.
Dr. Luo said previous breakthroughs in treatments and cures have relied on some luck.
"I would characterize previous successes in medicine more like a serendipitous event," Dr. Luo said. "You came across the findings and sort of hit the jackpot."
He said AI brings a huge shift to this space, allowing scientists to do less "wandering" and more discovering.
"It's sort of like a GPS," Dr. Luo said. "It will take us to the neighborhood we are looking for, and also along the way advance our existing knowledge of the human biology."
The idea of teaching machines the rules and characteristics of molecules or proteins is built on the same premise as the Generative AI that power chatbots like Chat GPT or Google's Bard.
"Something that people don't think about is, these large language models: they work for any language," Brian Martin, Head of AI for pharmaceutical giant Abbvie said. "If we think of DNA as the language of biology or protein sequences...a lot of these generative language models have a potential in those spaces."
Martin and his team focus on building machine tools that free up time for scientists to make more important decisions about human health outcomes. He said that process starts with a question.
"How do we accelerate scale and amplify what our scientists are doing already?" Martin said.
Martin showed NBC Chicago two imaging tools that he said can significantly reduce the time Abbvie researchers spend on repetitive tasks. One of the tools works with images of crystallography, which is an early part of finding a new drug.
"If we can automate [this process] from weeks or hours down to seconds, that’s time that goes back to the scientist and that accelerates and helps us contract that entire process," Martin said.
Another tool instantly counts the amount of cells in a tissue sample, and even recognize their shape.
"Imagine scientists having to spend hours going into and selecting and counting individual cells within an image," Martin said. "Those images are 200 times the size of the image that a normal person would have to use off of their cell phone."
While AI rapidly advance the capability to help humans, regulation of these tools is still being debated. Pharmaceutical drug discovery is already a regulated space, with governments around the world testing for safety and efficacy, but more rules may be needed.
"I think we should lay a solid foundation so this thing doesn’t get out of control once we start using it," Dr. Luo said.
AI models asked to generate beneficial substances can also generate toxic ones, but actually creating those materials once they're identified is another task entirely. Gamson and Martin both said a better future relies on keeping humans in charge.
"The same men and women who are responsible for the regulatory process that ensures safety and efficacy, those roles don't change," Gamson said.
In an industry designed to help -- and profit from -- important discoveries, serving patients first will be the key to trusting AI with the future of medicine.
"In our case, we act with a framework of rigor, respect and responsibility," Martin said. "And that helps define making decisions not really about whether we can do something but whether we should."
