Curing cancer is an iceberg more enormous than anyone could ever imagine.
In this episode, Pamela Bush, Chief Business Officer at Predictive Oncology, talks about her work connecting science and business to advance the discovery and development of drugs against cancer with the help of predictive AI and ML technology and the most extensive library of tumor samples.
Tune in to learn more about how Predictive Oncology impacts drug development!
Welcome to the Chalk Talk Jim Podcast, where we explore insights into healthcare that help uncover new opportunities for growth and success. I'm your host, Jim Jordan.
This is Jim Jordan and this is the Chalk Talk Jim podcast. Today's healthcare systems are changing quickly as they integrate with the broader economy, and that really has caused the necessity of this podcast to stay ahead of these changes, clinical, administrative, and business talent needs to access a broader understanding and insights into our healthcare system, and join us as our guests will help us gain these insights. So Pam, give me a little background about yourself and where in the healthcare system have you played, because I know you've played in several different areas.
Thank you, Jim, for the invitation, pleasure to be here. So I am a scientist by training. After completing my PhD, I went to the business side because I realized that I would be able to build bridges between scientists and have those conversations that are important to make sure that we advance healthcare and the life sciences industry. I have also worked in the pharmaceutical industry, in the finance side, as well as in the corporate business development side. And what that means is, as pharma is looking to complement the pipeline that they have of drugs that they are developing, sometimes they have to go to the outside to look for technologies that are going to complement and supplement the pipelines that they're developing. So I have worked with companies, with pharma, to make sure that they are finding the right technologies that they need. In addition to that, I have also worked in the payer side of pharma. Pharmaceutical companies have to work closely with payers to make sure that patients are able to have access to the drugs.
So without insurance companies paying for the drug, patients cannot get access to it. So it seems like you have a broad experience.
So I have worked from the research side through corporate business development, making sure that we have the right pipeline, the compounds or technologies being developed, and then all the way to the payer side and interacting with payer.
What today are you doing?
So today I am working for a small company that has the technology to help biopharma develop better oncology drugs. We all know that it is really hard to develop a new drug. A lot of the drugs fail when they are in clinical trials and this happens after years that biopharma has already been working on a particular drug. So there's a lot that happens behind the scenes before it even gets into humans, and then drugs go into human clinical trials and then they fail and they fail at a very high rate. For easiness of calculation, I'd say that about one out of ten drugs make it, which means that nine drugs actually failed in clinical trials.
Why do they fail?
And that is one of the areas that we're trying to address in the company that I work for right now. We are focused specifically on oncology, and there's a couple of reasons why they fail, but we address two of them, and one of the reasons that they fail is that until the point that drug goes into the clinic, those drugs have not been tested in a diverse population. So as you can imagine, 100 people that have a particular type of cancer, they're all going to be slightly different, they are going to be diverse, and that diversity is not included when researchers are doing the pre-clinical development because those metals are not set up for testing that diversity. And that is actually called heterogeneity and that is the variability that exists between human-to-human and that's also seen in tumors. So that tumor heterogeneity is not included in the preclinical stage because being able to access those tumor samples is costly, takes a lot of time, and right now the way that oncology clinical drug development is done, it is done with maybe a couple of tumor samples, it is done with immortalized cell lines, which is not the same as tumor sample, and it is done in animal models because that's the way that we have been able to develop the best system to screen and to test different drugs. The company that I work for, it's called Predictive Oncology. What we do is we have a platform that includes a very large number of tumor samples, it's a biorepository, and it's the largest privately held collection of tumor samples, and that is paired with artificial intelligence and machine learning technology that actually came out of Carnegie Mellon University. And what this does is it allows biopharma to test hundreds of compounds in the preclinical stage against hundreds of tumor samples and be able to predict the drug response of the different potential compounds in a heterogeneous, in a very diverse population of tumors. So that it is taking that insight that you would see not until clinical trials, you're able to have that insight about the response of your particular drug to a diverse population in a preclinical stage. So the drugs that are being moved into the clinical stage are going to have a higher probability of success because they've already been tested against dozens, hundreds of tumor samples.
So your goal would be, instead of one of ten making it, if you could make five out of ten drugs make it, that it would bring down the cost dramatically. You, too, just talked about the two challenges that your company solves, and it seems like the first challenge is the basic research suggests that a particular drug may help a patient. However, their data is not diverse, and by having such an extensive tumor sample library, you can test for that diversity.
So all of those cells, the composition of a tumor is very unique to the patient and the way that particular human being is going to respond to the drug is going to be very particular to that person. Even though we're creating hundreds of tumors, we are able to screen and predict how a particular drug is going to respond to a multitude of tumors, each person is going to be different. So those cells are unique to the patient, and what we're trying to do is by utilizing this platform, the complete tumor of a patient, we have all of those cells. We're able to then assess a, have a complete view as to all of the cells that would have been in that patient in the tumor to see how the whole tumor would respond to a particular drug.
So the second piece that your company is adding is artificial intelligence, and I should note for our guests that your company not only employs traditional biologists and chemists, but top-tier computer scientists as well. So everyone talks about artificial intelligence, it's thrown around like we actually all know what it means. So would you be able to describe the process?
Yes, and I will do my best because artificial intelligence is not my area of expertise, but I will do my best. So artificial intelligence, in this case, is the technology that is actually powering our ability to screen hundreds of drugs against hundreds of tumors. It is able to incorporate what a human being would not be able to process at a time. So it is incorporating not only hundreds or thousands of features about a particular drug, it's comparing that against thousands of data points that are coming from the tissues, and it is also incorporating any public information that is available about particular molecules or similarities between those molecules. So in this case, artificial intelligence is the ability of putting together an enormous and artificial, because no human would be able to process all of these kinds of permutations at a time, and being able to make calculations that as humans we would not be able to.
So I'm looking for an analogy here. So would this be like having the entire Google Map database and comparing a pothole in Massachusetts to a pothole in California and using weather as a variable to predict the growth of the pothole?
Yes, based on the weather conditions that we've had here and compared to what we have seen in that pothole that we already know, a pothole may have been born, absolutely, yes. So this is what the technology is doing, and it is absolutely, I find it fascinating. If you have ever played a battleship, your numbers 1 through 10, and then you have your letters A through, I don't remember what the last letter is in Battleship, but you have a two-dimensional grid. And on one side you have a list of all of the drugs that you want to test, all of the compounds that you want to test, and on the other axis, you have all of the tumor samples that you are going to be testing. You input into the artificial intelligence, all of the data that you have about the structure, the molecular weight, everything you have about those drugs, and then you also input all of the data that you have about those tumors. We've had these samples for over a dozen years, so there's a lot of drug response information on them, we know how these tumors have responded to other drugs in the past, so all of that information is put into the artificial intelligence. The AI core is going to make a prediction about every single combination of your drugs and your tumors. So imagine if you have a hundred drugs and a hundred tumors, you're going to have ten thousand different permutations. So the artificial intelligence can do that and it can predict, based on information that is available, and by the way, it's also pulling on public information, whether each of these combinations is going to be a response or not response.
Pam, you've given us a lot here, and let me just pause for a minute to make sure I have this right. So your company has one of the largest tumor libraries in the world, and your company expands this library by adding all available public knowledge and all known data from all FDA-approved drugs. Then you add all of this data into an artificial intelligence system so that you can move it from two- dimensional data, as in your battleship game analogy that we used earlier, into something multi-dimensional. With this information, what is the decision you're trying to make?
Based on the predictions that the artificial intelligence has made, we are able to look at like, what are the different data points that it's the least confident in. So because we have access to those tumor samples, we're able to go into the lab and we're going to be able to test drug A with tumor sample 37 and so on. We're going to be able to test those, get actual wet lab results of whether those tumor samples responded to a particular drug or not, and then put that information back into the artificial intelligence.
May I interrupt and ask you what you mean by wet lab?
Excellent, thank you for asking me that. The wet lab is when you're able to go into an actual laboratory and you're able to take a plate from which you're putting a small fraction of those tumor cells, and you're able to then test them with a particular drug and see how it behaves.
Thank you for that definition. So it seems like the AI system makes a prediction. Is it that your wet lab results can affirm or challenge the AI prediction?
So you see if the tumor actually responds to the drug, meaning the cells are dying or whether it doesn't. And then those results of yes, it responded to the drug, that is that cells died, or versus it did not respond to the drug and the cells continued the drug, that is all done in a wet lab, and wet because it's all with solutions and it's all with life cells. So then we take those results and we put it back into the artificial intelligence, we feed that information, with that new information, it is able to then re-predict everything. So it is able to make new predictions for all 10,000 of data points, and some of them are going to remain the same and some of them are going to change based on the new information they just learned. So it goes and predicts everything again, and then again, it gives you the least clumping, and we go around and around, and these are different rounds, as you can imagine, till it stabilizes, the system stabilizes and it no longer needs more information to be confident on the predictions that it has made of whether a drug and a tissue are going to be a response or it's going to be a nonresponse.
And this is amazing cutting edge. Clearly, this was not part of your PhD training program. Can you share a time in your career where you've had to quickly adapt or adjust yourself to keep progressing?
Oh, I've had to do that many times. So I came from academia into the business world looking to help companies get started, and I had to surround myself with people that were able to, and that wanted to, coach me and teach me about how science works in a business environment and what is the way decisions are made, and learned as how I can actually be able to talk science at a level that an individual without a science background can understand it, can distill that essence of the importance of the science, and the same way backward, explain why business decisions need to be made in a way that then the scientists are going to be able to adjust investments, why they're making the scientific decisions. Because ultimately, what we're all trying to do is make good investments in good science so that we can help people live better lives. So I think that there's value in being able to go back and forth.
Knowing your background as a PhD, you're a senior program director for a nonprofit Venture Capital Fund, and I think there you gain some great insight on investors' view of science. I think this resulted in you getting your MBA, if I recall, at CMU Tepper's Business School. After that, you went into a big company where you tend to move into positions that are much narrower in scope. I suspect that you actually had a broader view of the world than perhaps your boss. How did you emotionally manage through this?
By having a lot of conversations with people sometimes outside of my function, and that is one of the things that I think that the bigger the company, the functions are a little bit more siloed. And I took that opportunity to learn about how the different silos work, because I think that that was going to satisfy my curiosity about how a bigger organization works. Like you said, I had worked with smaller companies and everything is done by one, maybe five people at most, and when you're in a bigger environment, sometimes some of the groups, some of the teams, some of the functions forget or do not take into consideration what other groups are doing. I would say that being a science background in a corporate environment in which I was first working in finance and then working in corporate business development was different than anybody else in my team, there was nobody else that would actually speak the science, but that allowed me to have very good conversations with the people that I was serving and the people who are my stakeholder. So an example, I was supporting in finance different groups within the R&D organization, I was able to have very good conversations with the scientists that are the leaders of those teams and understand what it was that they were looking to do, and making sure that when I was bringing that message of what were their priorities to finance, that then we were able to shape the budget and shape the forecast to ensure that we had earned them with what they needed, and the other way around, just explaining then what were the restrictions or the parameters and maybe some objectives that we needed to meet within Finance, being able to explain that to the scientists so that then we could create a collaborative environment in which we were supporting the right science to advance. I think that it was through a lot of conversations and trying to meet as many people as possible to make sure that I was creating those bridges and those links, because ultimately that's what I wanted to do. I wanted to make sure that the best science advanced.
With all the adapting that you've done, where do you go to stay current?
So a whole lot of places and it has been reading some peer-reviewed journals. As I shared with you earlier, I am not an expert in artificial intelligence and I am nowhere close to that. So reading a lot of publications that have been put together from artificial intelligence, so through website searches, through some of our scientific advisory board recommendations, some of the foundational documents that became then the technology that we ended up licensing from Carnegie Mellon University to create a platform. So I would say I am a student of wherever it is coming from and just talking to people and asking questions and saying, hey, can you give me some advice on some publication or some article that would help me learn a little bit more?
So now that you've gone through this journey, what is your one lesson that you've learned, or something that's unique that you'd love to share?
That I truly was hoping that by this point in 2022, we would have already cured cancer. I've shared this with you a long time ago, and it continues to be my number one North Star is finding a cure for cancer, finding the right drugs to advance that have a higher chance of making it to the clinic so that they eventually get to the patient. What I have learned is that it is going to take a lot of effort and it is going to take a lot of people and it is going to take a lot of resources. So I think that the one thing that I'm learning is that the iceberg is bigger than I would have ever imagined it was when I was 14, and I thought at that point that that was going to be my mission in life. So I just keep going down and down the iceberg and it keeps getting bigger and bigger, but it is an iceberg and we will eventually figure it out.
So given that curing cancer is your North Star, why did you move from the pharmaceutical product side to the drug discovery side?
I honestly feel that any opportunity that we have as an industry to increase the probability at any stage of either discovery or development of a drug to learn a little bit more about it so that we can increase those chances of success later on, I think that that is where the biggest opportunity is and that's why I am such a big enthusiast and fan of what we are looking to do at Predictive Oncology of bringing that insight that won't come until drugs goes into the clinic, bringing that insight into the discovery stage. And here what I'm distinguishing is discovery is what happens before it goes into the clinic, into humans, and then development, I'm calling what happens after it goes into the clinic. So that insight of bringing that population diversity, that tumor heterogeneity that we're talking about earlier, if we can bring any of that insight into that discovery pre-clinical stages, that is going to have a huge payoff for knowing more about the drugs, understanding how they're going to behave once they actually get to the clinic. And our goal is to shorten the time of development, which of course, you can imagine, has financial implications as well. But the idea is, let's make smart decisions about where we invest and where we move molecules into clinical development so that then when they're in the clinic, they have the higher chances of being successful and getting commercial.
That's an awesome life purpose. Anything else you'd like to share?
I want to thank you for the opportunity to come here to your show, to share with you a little bit about myself and the really exciting work that we are doing at Predictive Oncology. I think that we have a chance to impact the way development is done, so I look forward to see what happens in the next few years and then be able to look back and say, this is where it started.
Thanks for tuning in to the Chalk Talk Jim podcast. For resources, show notes, and ways to get in touch, visit us at ChalkTalkJim.com.
Automated transcription by Sonix