Finding the right medicine to treat conditions like anxiety and depression can be tricky. A doctor will start you on one medication that’s typically well-tolerated and effective, but it could do nothing for you — or have terrible side effects. Sometimes it takes months of trial and error to find something that works.
It’s an incredibly common issue. Dr. Priscilla Chan told an audience at South by Southwest Wednesday that it could be streamlined if doctors could check drugs against a generative AI model of your cells and systems. Chan, who co-founded the Chan Zuckerberg Initiative with her husband, Meta founder and CEO Mark Zuckerberg, said using AI could be the next big leap for biomedical research.
“The hope is with those models we’ll be able to answer some of the hardest questions in biology,” Chan said.
Artificial intelligence has been a hot topic for just about everyone since its breakout moment with the debut of the ChatGPT AI chatbot in late 2022. This week, it was a major focus at SXSW in Austin, Texas, with conversations around trust, accountability and the future of work.
Last year, two scientists in Google’s DeepMind AI unit won the Nobel Prize in chemistry for their work using AI to predict the structure of proteins.
As for how this technology can advance science and medicine, it could take years, if not decades. And these AI models will likely just speed up actual lab research, not replace it. But Chan sees a world of possibilities.
What we don’t know about ourselves
Chan, a pediatrician, said much of how the human body works still eludes the understanding of science. Sure, it has been a couple of decades since researchers cracked the human genome, but genetics offers just a roadmap. Chan used the analogy of a Lego kit of the Millennium Falcon from Star Wars — the genetic code is the instruction packet. However, we still don’t know how the individual pieces come together to form the spaceship. And when one part doesn’t seem to fit right, that’s where medicine needs to step in.
Beyond the gaps in scientific knowledge about biology, we also have a limited understanding of how biology works within individual people. Based on a small number of samples, we have extrapolations about how the body is supposed to work, but that is a tiny dataset that doesn’t come close to representing the sheer diversity of humanity.
An AI model could help describe what is happening in one individual’s cells — personalizing medicine so that your treatment differs from mine.
“If we build the right data and AI models, we can better understand specifically what is making us healthy and what is making us sick,” Chan said.
Can AI speed up biomedical research?
Current research techniques are also slow and expensive in developing new drugs and treatments. Ideas have to be tested in a physical laboratory setting, which takes a tremendous amount of time and resources.
Chan doesn’t suggest eliminating the existing physical “wet laboratory” research. But a machine learning model — a hallmark of AI — can help identify drug candidates with a higher probability of working, meaning it might take fewer real-world tests to reach a workable solution.
The models won’t always be correct. They’ll offer solutions and ideas that don’t work out, maybe physically impossible ideas, but that’s why there needs to be a filter of real human scientists tackling the ideas a model produces.
“It’s not going to give us the full answer,” Chan said. “I don’t want you to think that scientists are just going to talk to a model and get all of the answers they need.”
The machines can help scientists find better questions, Chan said. “It’s going to be the hypothesis generator,” she said.
While many companies and researchers are looking at ways to use AI in hospitals and the treatment of patients, Chan’s focus is on advancing the basic biological research that makes future advances possible. She sees AI as a potential major leap for science, akin to the invention of the microscope, the X-ray, the MRI or the sequencing of the human genome.
“Health and medicine, it moves in leaps,” she said. “There are decades when research gets stuck, and then someone invents a new technology that completely changes how we see the human body.”
Read the full article here