We use both our own and third-party cookies for statistical purposes and to improve our services. If you continue to browse, we consider that you accept the use of these.
 In Advanced Applications, Cofactor Genomics, Molecular Diagnostics, Q&A

The Path to Deliver Improved Decision Making Tools for Patients and Physicians

In our last interview, we spoke with Erica Barnell of Geneoscopy about the use of data analytics to interpret the massive amounts of genomic data that have been collected over the last couple of decades. We’re continuing to learn from experts across the world of oncology, immunology, and genomics precision medicine to get an update and fresh perspective. In this week’s episode, David Shifrin is speaking with Dr. Bonnie LaFleur, Research Professor of Biostatistics at the University of Arizona BIO5 Institute. You can also listen to the entire recording on Soundcloud with Dr. Bonnie LeFleur here

David Shifrin: Bonnie, it’s great to have you as our guest and welcome. Please give us a quick overview of who you are, what you do, and how you ended up at the University of Arizona. 

Bonnie LeFleur: Of course. I’m trained as a statistician and most of my career has been in biomarkers. For many years I was in the biomarker discovery arena and also the mechanistic uses of biomarkers. Slowly over time, I started getting more into the translational side. When I came to the University of Arizona, I was working with a gentleman, Gene Gerner, and he had a chemo prevention drug that is now a pharmaceutical company located here in Tucson. We recognized that we didn’t really have any precision tools for helping patients make decisions to take a chemo prevention drug.

This got me thinking about the next step in my career, which was more translational, like we all talk about the bench to bedside. I actually left the University of Arizona and spent seven years in industry, particularly working on companion diagnostics and co-developing biomarkers along with therapeutic agents. My role there was diverse. I started off just being a functional leader of a group of statisticians and data management personnel to help promote these collaborations between device companies and pharmaceutical companies. We worked very diligently on the anti PD-1 and anti PDL-1 space, which was fantastic.

It was fast paced and really changed the landscape in regulatory science. I had an opportunity to come back to the University of Arizona, and now I’m in the Bio5 Institute, which brings scientists across many disciplines together. We focus primarily on pharmacy, science, agriculture, engineering and medicine. One of the initiatives is to help commercialize new technologies, diagnostics and treatments. I came back to academia, but with a different kind of perspective on how we can share these discovery tools with larger companies.

David Shifrin: I’m speaking on behalf of Cofactor Genomics here, thinking a lot about immune-oncology, but is just one piece of this larger puzzle. Can you elaborate about that perspective and how these different areas of research, physiology, and molecular biology affect each other?

Bonnie LeFleur: Sure. Remember, I am approaching this from more of a data perspective. My colleagues over time have included oncologists, but also immunologists. Early on in my career when we were both at Vanderbilt coincidentally, I worked a lot with the immunology group, specifically pediatric immunology. We did a lot of work on the development of the innate immune system in babies, and looked at things like cord blood because that’s where there’s a shared immune system between the mothers and the babies. Then I worked in oncology for a very long time and looked at some of these novel immune therapies. Now I’m working with the Immune Biology Department here at the University of Arizona and we’re interested in profiles of aging and immune aging. I’ve also worked with patients that may have perturbed immune systems, for example people with HIV that have been on heart therapy for a long time. So I’ve worked with a spectrum of the immune profile throughout life.

David Shifrin: Well, it’s funny because that’s essentially a perfect encapsulation of what we’re taught in college or grad school. All the professors say development and cancer are just different versions of the same processes. We don’t even think about it anymore because it’s been ingrained in us for so long, but that’s exactly what you’re describing on a very granular level.

Bonnie LeFleur: Early on when we were looking at mechanisms in development, we didn’t have as many tools as we have now. Technology has just exploded since we both started working in the field and we’re able to measure so many different aspects, which is amazing.

David Shifrin: Let’s talk about both some of the opportunities and challenges. When looking at the immune system when it is perturbed, how do we identify biomarkers? 

Bonnie LeFleur: I’m coming from a data side and I’m very interested in helping people make decisions. The first people we talk about are patients, and when they are afforded an opportunity for a treatment, we can provide decision making tools for a risk benefit. I view a lot of the assays and measurement technologies as things that give more and more information to the patient so that they can make those risk benefit decisions. 

I think we’re doing a great job on developing assays to evaluate a single point in time, including genomic and proteomic markers from tumor sections using RNA-seq based and quantitative RT-PCR based tools. We’ve optimized the guidance for assay validation. When I was in industry, it was the first time that these kinds of guidance had been put together in such a manner to co-develop a biomarker and therapeutic agent. We have a path for that now that wasn’t there 10 or 15 years ago. I think these are really helping us understand how to help patients make decisions.

We also have a lot of standard operating procedures for platform comparison studies. I’m referencing the MAQC, which consisted of microarray and sequencing quality control projects that were initiated through the FDA early on. They started with microarrays because we needed a common standard practice with respect to quality control. Now they’re in the MAQC-IV, which is more clinical and answers the question of how do you take RNA-seq technologies and use them in clinical practice? They give people experimental designs for comparing different platforms, which for example, helps us when we want to do things like a blueprint study.

When we compared all of the PD-1 and PDL-1 by IHC tests, the blueprint didn’t follow exactly this platform comparison, but they did use a common standard practice for comparison. Even more recently, is the circulating tumor cell comparison. It was an orthogonal comparison of plasma and next generation tests for circulating tumor. Those kinds of SOP’s are revolutionizing our ability to evaluate how well these competitive tests and assays perform, how we can harmonize them, and also how we can identify what’s working and what isn’t working in standardized way.

David Shifrin: Anytime you can standardize and benchmark, especially when you’re talking about regulatory and intellectual property, it’s incredibly helpful to compare against what came before. 

Bonnie LeFleur: This helps us because oftentimes as a statistician, helping people understand a common standard or experimental design is critical. This is especially helpful when you’re doing a platform or assy comparison study, as is identifying what the metric is that you’re comparing. This is incredibly relevant in immune oncology, and happening at the same time as this explosion of precision medicine opportunities. We really need a path for combining a lot of different technologies. For example, both kinds of IHC, high throughput, proteomics genomics, and metabolomics. These things seem to be working together to create this environment of health related issues, many of which are immune driven.

David Shifrin: I’d love to look at that a bit more. The idea of combining this multidimensional set of data not just in a static point in time like a slice for IHC – but progressively when looking at any given tumor sample or any given patient. You’re a statistician looking at the numbers and thinking about how to do this in a way that is useful, reproducible, standardized and actionable. How do you strip away the hype and get down to the numbers to make it useful?

Bonnie LeFleur: That’s one of the larger challenges that we have. From my perspective, once we have a path, we think about tools and that’s when we see just how complicated it really is. What everyone is seeking is something that might be more quantitative. Something that could provide a score or a decision path that helps people weigh risks and benefits based on a whole plethora of information that we have about the patient. Although that sounds really awesome, there are so many things we don’t know how to do, including combining all of this multi-sourced data to find truly multivariate patterns. As one feature changes, something else changes somewhere else. This is especially true in the immune system where there are so many pathways. If we block one, our immune system just finds another way around it.

David Shifrin: That description that you provided almost makes it sound linear, but it’s not. There’s a branching network of effects and they probably loop back on each other. It’s crazy.

Bonnie LeFleur: The good news is that there are a lot of people doing work in cyber infrastructure, which can harness cognitive computing to perform highly dimensional analyses using newer algorithms. More importantly, it gives people a place to share ideas. Without collaboration, it’s going to be very difficult for us to answer all of these questions. Cyber infrastructure is a platform for high throughput computing and people sharing tools and data with different perspectives. 

David Shifrin: I’m talking on behalf of Cofactor Genomics, which is a company doing their own proprietary work, but the people listening are clinicians, academic researchers, and possibly some patients. We’ve bounced back and forth between academia and industry, so how do you see all of these different sectors and stakeholders working in this larger cyber collaboration?

Bonnie LeFleur: We already have a nice network of key opinion leaders in the initial strategy sessions for the industry. A lot of times, those collaborations involve real world data. What could happen when we start combining tools and different technologies that are measuring different things like the genomic, proteomic, metabolomic, and clinical information?

How do we harness all that information to help direct the actionable treatment strategies? It’s often not just a clinical trial, single site, or even a small multisite study. That’s a really important aspect of the immune-oncology field right now. This is demonstrated in the Foundation Medicine paper that was just published with Flatiron where they talk about how can we harness some of these real-world evidence studies in a way that can be used to support and push forward clinical research.

David Shifrin: One of the things that we’ve sort of danced around but haven’t talked about explicitly, is the role of machine learning. It’s not just static data points from different samples, but multidimensional and includes real time data as well. What are your thoughts?

Bonnie LeFleur: While most of them are actually statistical tools, they’ve been modified to handle multiple kinds of data, maybe an unsupervised way. The statistical part is the harnessing of that information that you might acquire from an unsupervised type of an algorithm. You still have to make sure that whatever you obtain from these tools make sense and is interpretable. I think machine learning algorithms are very useful, but my job is to take these outputs and interpret how can this data inform a decision. How do we give the patient or company the tools they need to make a decision?

David Shifrin: Machine learning, like big data or the omics, has reached this fever pitch in terms of buzzwordiness. What is it really, and how do we use them?

Bonnie LeFleur: The pipelines are very complex and require a high level of quantitative expertise that’s not easily condensed into algorithms. Having an algorithm is great, but what does it mean clinically? How do you interpret the output?

That’s a challenge that we would all like to solve. When we try to understand the immune system and immune oncology, how do we harness that without pervading other positive parts of the immune system? There’s an enormous amount of knowledge and expertise required for not just the analysis part, but also processing the samples. Everything from selection of the section in the FFPE sample, all the way through the extraction, purification, etc. – all of these things are expert driven. There’s a level of person-level expertise that’s required in order to build these in a way that can be actionable. That’s a challenge for all of us because as much as having everything automated would be beneficial, at least right now, it’s not possible.

David Shifrin: What are a few areas that you think hold the highest promise? I’d especially be interested in any specific biomarkers or forms of analysis that you bet on heavily.

Bonnie LeFleur: I am really excited about the tumor microenvironment. A lot of my colleagues are interested in the 3D spatial organization of the cells related to their microenvironment. I think this makes the most sense. If I were to put my money on anything, I would need to put my money on something that’s going to allow us to monitor patients over time, postsurgical removal. Finding ways that we can measure this microenvironment including the inflammatory cytokines and T and B cell markers is critical, because we need to be able to monitor how well the current therapies are working and detect when they stop working. I also believe that there’s a lot of information in adverse reactions while on some of these checkpoint inhibitors that could be due to some subclinical disease that wasn’t found prior to therapy. We can use those risk biomarkers longitudinally, even in our patients that are likely to respond to treatment because the biomarkers could be used as a surrogate for the start of the response. 

David Shifrin: As you’re talking about this, I had a distant family member comes to mind who is participating in an immunotherapy clinical trial for advanced lung cancer. He got word two days ago that all of his scans were clear, including the primary site. I’ve been thinking for the last couple of days – It looks incredible and it’s amazing, but will there be a relapse? What are the molecular pathways? What is his immune system doing to create this response? At his age, where is he going to be in 5, 10, or 15 months?

Bonnie LeFleur: Right, and we don’t necessarily need to wait. We need something that would be able to monitor these patients that have responded very well over time. Then, we could catch delays in response or when they quit responding because the immune system will eventually find a way around an inhibitor. We’d like to find that prior to a recurrence.

David Shifrin: Got It. Well, Bonnie, this has been a lot of fun. You’ve been incredibly generous with your time, so thank you so much. 

Bonnie LeFleur:  Thank you very much.


Recommended Posts

Start typing and press Enter to search