How AI is aiding cancer research

PORTLAND, Ore. — At OHSU’s Knight Cancer Institute, researchers are focused on finding ways to detect cancer earlier and produce more targeted treatments for individual types of cancer. But with so much variety in cancers, let alone the human body, the amounts of data they need to gather and analyze is mind-bogglingly vast.

That’s where advancements in tech like artificial intelligence are coming in clutch — helping researchers not only compile all that data, but potentially to go through it all and find patterns that researchers can use to help cancer patients.

Dr. Brian Druker is one such cancer researcher. He’s also CEO of the Knight Cancer Institute, funded through a billion-dollar endowment sourced from Phil and Penny Knight, taxpayers and thousands of people who donated to support the effort back in 2015.

When The Story’s Pat Dooris sat down with Druker recently, he asked about how that funding works now that the institute is fully up and running.

Brian Druker: Let me start by thanking the 10,000 people who made this possible by donating and matching Phil and Penny Knight $500 million to make a billion to invest in cancer research. And that’s a huge responsibility, and we’ve taken that seriously and we’re breaking new barriers, new boundaries in what we call precision oncology — which is matching patients with the right treatment — and early detection of cancer so we can identify cancer at its early most curable stage. And we’re making significant advances in both those areas as we promised.

Pat Dooris: OK, we’ll get to that in a sec. But, first, I think some folks at home may wonder what do you do with a billion dollars? I mean, does that go into an investment account, or what do you do with that? 

BD: First of all, it’s being invested over a decade, so it’s not like there’s a pile of gold sitting somewhere. But we put about $250 million into an endowment that will give off operating revenue year after year over a decade, so we’re sustainable. But mostly what we put that into has been people and programs: $200 million came from the state of Oregon and bonding authority that built the building that now houses over 875 people who brought in the last decade over $1 billion in grant funding from external sources to the state of Oregon. But then we’ve recruited incredible talent, people to come and work on curing cancer here in Oregon. And that’s been the best part of my job is to bring that talent here and watch them create.

PART ONE: Top doc at OHSU’s Knight Cancer Institute is on the bleeding edge of research

The building Druker referred to is the Knight Cancer Research Building, which is also where the interview took place. After it was finished, Druker went on a recruiting binge — building a team that now numbers 1,500 people, all contributing in some way to the most advanced cancer research in the world.

It’s a monumental effort, which brings along with it a certain amount of pressure to produce results. While much of their research is a work in progress, there have indeed been results.

BD: Well, a couple of big breakthroughs: Not only do we have the ability to target cancer, but we’ve also learned how to harness the power of the immune system. And although we weren’t the center that discovered this, we certainly are using an enormous amount of the immunotherapeutics. But an investigator working at the time in New York determined that when we develop cancer … the cancer tells our immune system to put the brakes on: ‘Don’t go after that, just leave it alone.’ We’ve learned how to take the brakes off, so your immune systems say ‘Hey, wait a second. that’s a cancer. We better go kill it.’ And by reactivating the immune system to do his job — to attack and kill cancer — we’ve had remarkably effective therapies.

One example that I give a lot of people is former President Jimmy Carter. He had metastatic melanoma to his brain. He’s still alive, celebrating his 100th birthday over five years after what should have been a fatal diagnosis in less than six months to a year. Those are the sorts of treatments — and now what we’re doing is we’re using our smart analytics, also trying to figure out what’s the immune system doing and can we help trigger the immune system in better ways and combine them with these targeted therapies to get even better outcomes. 

PD: And I know you’re talking to the 10,000-foot level and I appreciate that, but is there — similar to the earlier question — a simple way you can explain how you’re able to tell that the cancer is turning off the immune system and you tell the immune system go get it? OK if it’s too complicated. 

BD: Basically, what we do is we can actually look at the state of an immune cell and if we do some very sophisticated profiling, we can determine whether it’s in what we call a resting or inactive state or an activated state. And we can actually look at the switch between those two states and determine, ‘OK, those immune cells are off. We treat, they’ve turned on, and they’re doing their job.’

Druker also points to artificial intelligence as a key development that will allow researchers to study the DNA sequences that make up our human genes, helping to find clues about what is happening with different cancers. It’s an exciting field that Druker and the Knight Cancer Institute are embracing.

PD: You’ve sort of already been talking about it, but what’s the future? Where do things go from here? 

BD: Well, they go a couple of places. I still think we can do better at matching patients with the right treatments and we’re working quite a bit with any number of companies in the artificial intelligence realm, and we’re actually hiring a large group, thanks to philanthropy, to establish what we call a Center for Biomedical Data Science in collaboration with our colleagues at the University of Oregon. Because we still think that there’s answers that cancer could give us if we knew how to analyze the data better; we can generate massive amounts of data. But I can sit down with the data and say, ‘I think this is what we need to do,’ and sometimes, it doesn’t work. Sometimes, it does.

Imagine if we had a computer that could tell us all the different algorithms that I can’t see and can say, ‘Why don’t you try this and maybe it’ll work better?’ So if you combine a computer with my brain and other really bright people, we think we can do even better. But at the same time, we’re also collaborating again with numbers of artificial intelligence companies looking at how does cancer go from early benign growth to a lethal malignancy? What does that trajectory look like, and how can we intervene?

And I’ll give you an example we’re doing in prostate cancer: We have 10,000 prostate cancer biopsies from our collaborators in the United Kingdom with clinical follow-up. If we ask a pathologist, the people that read these biopsies to grade those 10,000 prostate cancers, it would take them years to look at those 10,000. We can train a computer to look at a type of early-stage prostate cancer, train it, and then, it can go through those 10,000 samples in minutes.

Then we can go back and train it on a slightly more advanced stage. And after we’ve trained it, it can go back and look for that in minutes. So now, instead of requiring years to do these analytics, it’s days to hours, days to minutes. And now we can start to say, ‘OK, correlate that with the outcome and tell us which men you predict are going to progress and which men we could just leave alone.’ Those are the sorts of things that we’re unlocking with these powerful analytical tools that we never had before.

There are plenty of potential dangers that come with the advent of artificial intelligence. But cancer research is one field where it offers clear benefits for humanity — and Druker thinks it’s incredible the way that AI allows researchers to leap forward with new discoveries.

BD: Well, let me give you an example. When I first started out in the lab a long time ago … I could, if I sequenced one gene in the human genome, one out of 30,000 genes, that was a 10-week project, and it was an amazing accomplishment. One gene out of 30,000. (In) 2001, the first human genome — all 30,000 genes — was sequenced at a cost of about $100 million. About 10 years ago, you could sequence the entire human genome for $10,000. Now, I can take a tumor, and I can sequence each individual cell — tens of thousands of cells, their entire genome — and with barcoding, I can identify each individual cell in that cancer. So now 30,000 cells (and) 30,000 genes. That’s massive amounts of data from one cancer.

PD: And is that like from a cell being fed into a computer because you’re not looking at it under a microscope?

BD: It’s being fed into a machine, and the machine spits out all that data. Now, our job is to try to analyze it and make sense of it. 

BD: Yeah, wow. Even for somebody like me to look at this technology, it’s — this is way beyond science fiction.

PD: It almost sounds like with AI, the only limit is your imagination for the way they’re used. 

BD: It’s imagination and data. You need a ton of data, and we’re collaborating with any number of people around the world to ensure we have enough data to train the models and then help us design the experiments that are going to make a big difference for people. 

PD: And as you were saying that I was thinking, I imagine the data generally has been out there, but maybe kept in one university or lab or it’s not been centralized? 

BD: Right. And I think what everybody realizes that by collaborating, by combining that data, we’re going to solve this problem so much faster now. One of the things we did almost a decade ago in another type of leukemia, of lethal leukemia called acute myeloid leukemia, is with support from (the) Leukemia and Lymphoma Society, we collaborated with about 12 to 15 different centers around the country, about 10 or 11 drug companies, and we profiled over 1,000 patient samples in incredible depth. And we’ve made all that data public. So, now, if somebody says, ‘Is my favorite gene or pathway activated?’ they can go search that in our data in our database, and instead of having to collect hundreds of samples over years, they can sit down at a computer console and do that experiment. 

PD: It’s fascinating and exciting to think about it. 

BD: It’s truly exciting times for cancer research. 

PD: Will there ever be a time where cancer is fully defeated? 

BD: You know, I think in my view, what I want to think about with cancer is that it’s no longer fear … I don’t think we’ll ever eliminate cancer — much like I used to say, I used analogies to infectious disease. I’ve stopped using them because when COVID came along, I was like, ‘Whoa,’ you know? But before COVID, you think about infections. What antibiotic do I need? Well, it’s a mild cold that will run its course. With cancer … when you hear you’ve got cancer, your mind is numb. I want it to be it’s ‘You’ve got cancer.’ ‘OK, well, what do I need to do to get on with my life?’ It’s not a big deal. For my patients who take Gleevec, that’s the way it is for them — and that’s what I want for many, many others. But ultimately, I want to make sure we’re detecting it early when it’s highly curable, and also, we need to think about prevention, where we actually … if you don’t get cancer, that’s always going to be better.