Pamela Becker, MD, PhD, Professor of Medicine at the University of Washington School of Medicine talks about using algorithms to fight cancer.
Interview conducted by Ivanhoe Broadcast News in July 2018.
Tell us a little bit about the Pick a Drug and how it works in the trial.
Dr. Becker: How it would work for a patient is that their physician would know that the patient is eligible because they had failed a few different types of treatment already. Basically they would have tried all the usual treatments for acute myeloid leukemia. That physician would contact myself and my team and find out if they would be eligible to enroll. Then my team would check through different aspects about them, like do they have any other illnesses that would make it harder for us to treat them. And if they have had enough tries at usual treatments, they would come by and we would arrange for a conference where we explain the procedures and they would sign a consent form. After they signed the consent form, we might go on to draw their blood if the leukemia cells are in their blood, or take their bone marrow if the leukemia cells are in the bone marrow, or both. Then we would arrange for those to be brought over to the laboratory at the South Lake Union Campus of UW Medicine. There, at my own laboratory, we would work on that blood or bone marrow and we would isolate the leukemia cells. We would then bring them to the Quellos high throughput core facility to spread out those cells in tiny wells in plates that are specially designed for robotic devices to add all the cells and then add all the drugs. So we have one hundred and fifty drugs and drug combinations. The cells stay on those plates for about 72 hours, and then we add an agent that shows if the cells are alive or dead. We check if the cells survived those drugs or didn’t survive those drugs and then we would find out which drugs might be the best for that patient. We then provide that information to the patient’s physician and suggest that which might be good drugs for their patient. That’s f an overview of how all that works.
And so the way you would pick from those one hundred and fifty drugs is from that algorithm?
Dr. Becker: The way we would pick those drugs is by choosing the one that had the highest drug sensitivity, meaning the one that was most effective at killing the leukemia cells. That’s how we would pick what we think would be a good drug for that patient. But with the new algorithm, what we are able to do now is take the information into consideration in terms of what genes the patient expresses. In addition, our protocol includes checking 194 genes for mutations in those genes that we know are associated with acute myeloid leukemia for which we might have effective drugs. The new algorithm connects gene expression with high thoroughbred drug sensitivity testing. All three components are going into the decision-making to pick a drug for the patients. So we have results of the high throughput functional screen within about five and a half days. We have the mutation analysis within about 10 to 12 days. On average, we’ve been picking drugs for patients in about eight days. So it’s actually incredibly fast for us to get all of these kinds of results. It’s taking us longer to get the gene expression data that would then be used to pick which drug classes might be good for that patient. That’s a longer-term process and we’re developing a brand-new clinical trial that will be our third such trial, currently in preparation. That one would incorporate this information on the gene expression and the algorithm would predict as well as it will also be testing much earlier in the course, at the time of diagnosis. We would start with patients with newly diagnosed disease and we would start their testing right off the bat to find out what the vulnerabilities would be for that individual patient from day one.
So the gene expression is the algorithmic part of that?
Dr. Becker: Exactly. The gene expression is how we match if we have genes that are highly expressed. The algorithm helps us to predict which drugs would be good for that patient, which drugs to which they might be sensitive, and which drugs to which they might be resistant. That’s how, and it’s a few dozen genes.
So, with a combination of all three inputs of information, it helps you narrow it down?
Dr. Becker: Exactly. Usually, for every patient we have many choices of drugs, which is important because many of the drugs we’ve tested are investigational. We have to find out if there is a clinical trial or if we can get the drug from the company. Otherwise, we have to go through the patients’ insurance companies and gain approval for the drug or the drug combinations. So having the different choices is actually very helpful. In the future, hopefully, we’ll be able to get more drugs for the patient. Right now, we may have to go down the list to their tenth choice or their twentieth choice, but we hope in the future to be able to go higher up the list of preferred agents because of better drug availability.
How has the addition of the algorithm improved effective success? Effective drug choice, good drug choice for doctors?
Dr. Becker: It’s a future event. So, because we just invented it, we haven’t actually gone live yet.
I got you.
Dr. Becker: It would be really hard for that to be happening all over the country just as of yet. I think that as these algorithms become more robust and as we develop more and more proof for them. Right now, we’re still in the testing mode. We still want to prove that every time that algorithm makes a prediction that it correlates with sensitivity in the laboratory, and then we eventually need to prove that the algorithm is working in patients. That’s going to take a few more rounds of clinical trials to prove that it works. Then it could go live and help people in real time.
From what you’ve done, I don’t know if theoretical is the right word, but do you have any information or feeling that this is something that’s really going to help?
Dr. Becker: So far, we took one of the predictions for a gene that, when highly expressed, conferred sensitivity to certain classes of drugs we use in AML. By over expressing that drug in cell lines, we proved that the prediction worked. So we showed that we could cause those cells to be more sensitive to those drugs by over expressing that gene. That’s how you do a proof in the laboratory for a theoretical question.
And then it moves on to animals and then people.
Dr. Becker: Yes.
In the big picture, this would cut down the clinical trial part of it because of your theory. Go ahead and comment on that.
Dr. Becker: Yeah. So how I envision the future of precision medicine is a little bit different. It’s not just that a patient has a mutation and therefore we can give an inhibitor for that particular protein to that patient. The reason is cancers have hundreds of mutations all at one time. I’m also looking for mutations in 194 genes at the same time, so it’s more comprehensive that we look for vulnerability in terms of mutations. We actually have found that in 80% of the patients we can offer a treatment based on those mutations, which was the case in both my first and second trial. We’re taking that mutation information, the gene expression information, and the drug sensitivity and laboratory information, all of which goes into databases which we then hope will be analyzed by perhaps even newer machine learning algorithms. Then what we hope to find are like fingerprints exhibited by the patients in terms of what their mutations are, include other characteristics such as age, other illnesses, and what their drug sensitivity shows. Using prior patients’ outcomes we can say, well every time a person shows up with that fingerprint they seem to respond best to this set of drugs. So the machine would associate them with that particular fingerprint and then we would offer the same kind of drugs that have been successful for similar patients. It’s how we analyze those big data. I imagine that we’ve tested almost a hundred patient samples thus far in our assay. So if we take one hundred patients and we have the mutations for 194 genes, we have their gene expression for 17,000 genes. We have their drug sensitivity to 150 drugs and drug combinations and you can envision how much information this represents,and how many different permutations exist. Ideally, after entering information from our tests, the machine algorithm would somehow spit out how we should treat that patient based on gathering how previous patients responded to all the drugs they did receive. We need that last piece, which is what happened in the patient. Then in the future, with more advanced machinery, we would be able to, I think, make great advances. We know that every single patient’s cancer is different from every other. So, to be able to take all the data that we have and even data that we aren’t gathering yet—there are other layers of additional data, proteomics, metabolomics, and other features that are not yet analyzed, but efforts are underway r to analyze that these other features. Then we could fully characterize every single cancer and be able to know exactly what the vulnerabilities, or Achilles’ heels exist for each patient. And notice what’s consistent amongst groups of patients and then be able to offer the best drugs. Starting with what would work best would also save patients from a lot of toxicity. For forty years we treated everyone with the same pair of two drugs, and only a certain fraction of those patients responded. And a much smaller faction of those patients survive after chemotherapy treatment and then a stem cell transplant from a donor. Only 25% of the patients are surviving long-term. That means that for 75% of those patients we need to find a better way.
So to get all of these things working, the gene mutation and everything else, we’re years away still from actually starting a test for patients?
Dr. Becker: Well, we already are starting to test in patients. Depending on the results of how they do with the drugs that we pick, we’ll then go back into the algorithm. So it’s an ongoing process.
So the algorithm is not being used in your Pick a Drug 2 yet?
Dr. Becker: We will—that information will be used in a future algorithm. It’s like continuous learning — it’s using more and more data and then building more and more hypotheses and then testing more of those hypotheses. So it’s an ongoing process. I would say that right now we’re going off of everything we do have. In the future we’ll have more, and then we’ll just keep building. I wouldn’t say that we aren’t treating patients yet. We are already treating patients because we know there are options. The patients that I’m treating now have exhausted all the usual possibilities. Many of them have already had a blood stem cell or bone marrow transplant. And so there is nothing else that can be offered to them other than Phase I trials where the drugs are just being tested for the first time in people. This enables us to actually try to use the information from their own cells and try to pick drugs that we think would work for them. So it’s a different way of approaching the treatment.
So for people with the cancer that’s not easily treated and normally doesn’t have a very good outcome, this has to bring a ton of hope.
Dr. Becker: Exactly. So it does. In fact, what I just described that we’re doing and how we are doing it, many patients have said to me why isn’t everyone doing that? But I realize that it was about ten years ago when we first conceived this approach. So it’s been a long time of working on it. And every step of the way there’s been great advances. But even the last two protocols that we’ve run have been called feasibility trials because I had to prove that you could actually do this in patients. That you could actually take a blood or bone morrow sample from a patient here at University of Washington Medical Center or Seattle Cancer Care Alliance and have it taken over to the core facility in our research building at South Lake Union and have it tested and get results. Then actuallyprovide the information to the doctor and find a drug that was available that we could give to the patient. Just to be able to prove that part of it has taken a long time. We’ve been doing these trials for several years now.
What haven’t I asked you about the algorithms or this precision method that you think is important to include in the story?
Dr. Becker: Well, I think that each patient is an individual and has a story to tell. The fact that we can give patients and future patients hope, I think that that’s the element that’s hard to come by through all the science. So you’re thinking about the computers and the machines and the data but realizing that there’s a human being behind that with hopes and aspirations and interests and you’re able to help them fulfill their dreams. That’s the part that we haven’t talked about yet.
That’s awesome. How has it helped you personally and professionally, this information?
Dr. Becker: I think it’s been a little too early to have made my job easier, but I would say that there’s enormous gratification when the science seems to work. When the patients send an email and say that a patient was able to attend the family members’ wedding, or a patient is still fighting their cancer and it’s almost a year later and we feel that you really helped us because we got to another transplant. Or we got to a subsequent therapy that was successful and now we want you to help us again. I think that there’s a lot of gratification in being able to see the small successes and being able to see the bigger picture. And to be able to find this correlation and to be able to have hope that I may not have to watch patients succumb to their disease in the future. So I think that I am fulfilling my own dreams in terms of being able to make a difference–I guess the question is what keeps me going every day. The science keeps me going and the patients keep me going. I feel that they go through a lot of suffering through all the chemotherapy treatments and the transplant procedures and they’re brave and they’re determined and so I feel like I should do everything I possibly can for them.
END OF INTERVIEW
This information is intended for additional research purposes only. It is not to be used as a prescription or advice from Ivanhoe Broadcast News, Inc. or any medical professional interviewed. Ivanhoe Broadcast News, Inc. assumes no responsibility for the depth or accuracy of physician statements. Procedures or medicines apply to different people and medical factors; always consult your physician on medical matters.
If you would like more information, please contact:
Bobbi Nodell
206-543-7129
Sign up for a free weekly e-mail on Medical Breakthroughs called
First to Know by clicking here.
