Tissue Phenomics: Closing the Gap Between Genomic Information and Patient Outcomes

Thomas P. Heydler

Interview with the Innovators

An Interview With Thomas P. Heydler, CEO of Definiens

Thomas P. Heydler is Chief Executive Officer of Definiens. Mr Heydler holds a Diplom-Ingenieur degree in Electric Engineering and Computer Science from the Technical University in Munich, Germany.

Perhaps the most engaging quality of personalized medicine is the astounding diversity of its facets, each of which changes how clinicians approach patient risk identification, profiling, diagnosis, treatment, and follow-up. Understanding the science of genomics essentially opened up the gateway to personalized medicine. More recently, the science of phenomics is capturing the attention of the biotech community, and it is this topic that constitutes the focus of the following interview with Thomas P. Heydler, CEO of Definiens, who outlines an emerging dimension of personalized medicine by explaining how the role of imaging is redefining the decision-making process by combining and correlating genomic information, tissue-imaging data, and clinical outcomes. Definiens holds that this is the means of achieving true personalized medicine in oncology. The avenue for obtaining this diagnostic information is Definiens’ Cognition Network Technology^®, which extracts meaningful, context-specific intelligence from many types of data, including text, numerical data, or images. The goal of this technology is to maximize the value of the data and ultimately aid in decision-making processes. The company’s current focus is on image analysis in life sciences and healthcare.

This technology captures more information contained in histological slides in order to develop novel treatments and bring those treatments to the right patient at the right time. Specifically, Definiens aims to support pharmaceutical companies in bringing targeted drugs to market faster, with smaller and more conclusive clinical trials. Recently, Definiens signed an agreement with Inserm Transfert to sponsor the worldwide consortium for the validation of the Immunoscore™ test, a new approach to cancer therapy. A team led by immunologist Professor JĂ©rĂ´me Galon demonstrated that the number, type, and location of tumor-infiltrating lymphocytes in primary tumors have prognostic value. Clinical data show that patients with a strong Immunoscore have a better prognosis and that immune response may be a better indicator of risk for recurrence than standard tumor staining. Definiens has developed the underlying technology, enabling the accurate and reproducible quantification of tumor-infiltrating lymphocytes essential to the Immunoscore test.

Under an international task force led by the Society for Immunotherapy of Cancer with the partnership of the European Academy of Tumor Immunology, the Cancer and Inflammation Program, the National Cancer Institute, the National Institutes of Health, and La Fondazione Melanoma, the Immunoscore test is being validated by immunologists and pathologists in 15 countries.

We recently spoke with the CEO of Definiens, Mr. Thomas Heydler, who graciously assented to not only explain Immunoscore technology and its role in reshaping personalized medicine in oncology, but to do so in terms relevant to the clinician’s needs and frames of reference. We agreed at the outset of our conversation that it would serve the interests of clinicians and patients to examine the new phenomics technology in practical terms consistent with practitioners’ needs and interests, avoiding a purely theoretical discussion.

One goal of this interchange was to understand how the infiltration of inflammatory and lymphocytic cells, as well as knowing their number, type, and location in primary tumors, provides oncologists with prognostic value in lung and colon cancers. Before diving directly into the science, it was useful to hear Mr. Heydler explain and position the core work and mission of Definiens. The ensuing dialogue achieved this mission of clarifying their first-in-class technology.

---

PMO Can you please describe the work that Definiens is doing and how it relates to personalized medicine?

Mr Heydler Definiens is a Tissue Phenomics™ company, which means that we believe there’s a tremendous amount of valuable phenomic information represented in tissue. This information is primarily used by pathologists to make educated decisions to advise oncologists about diagnosis and treatment. Unlike the field of genomics, the challenge facing phenomics is the absence of technologies allowing extraction of information from tissues to help oncologists make better diagnostic and therapeutic decisions. At Definiens, we have made it our objective to fill that void with a process for extracting all the information from tissue in order to understand not only the objects in each cell, but also the relationships of those objects within the tissue. In addition to fully quantifying tissue, we also understand the morphology, and use that information to create a fixed data readout of the tissue that correlates that information with other data sources, such as patient outcomes data. One application of this is to see which markers show the highest correlation between the cancerous tissue and progression-free survival. This combination will help us discover either new prognostic or predictive markers, which are the foundation for tests and then for better diagnostic services.

We are not calling it tissue sequencing, but tissue datafication. And we are hoping to extract all the information we can from the tissue to drive better diagnostic and therapeutic decisions.

PMO That puts it very clearly into focus. It speaks volumes. So tell me, how do oncologists interpret Immunoscore findings to select and titrate drugs differently than they would using the genomic approach, and how does this translate into improved prognosis?

Mr Heydler There’s a big gap between genomic information and patient outcomes. Tissue Phenomics is closing this gap. This is all about how a specific gene is translated into a specific protein, and that protein’s expression in the tissue structure is ultimately the manifestation of a certain disease.

In the Immunoscore field, it’s all about how fit the immune system is. Immunoscore gives the oncologist a score assessing the strength of the patient’s immune system. In order to do this, you have to look to the tissue. Your question is well put, because you’re not looking only to pick a marker and identify the expression of specific markers in the tissue, but more importantly, it’s all about morphology. It’s all about the correlation of lymphocytes and T cells, for example, toward the cancer cells and how those immune cells are grouping in certain areas, such as the core tumor or in the invasive margin. That’s the key indicator on top of the expression of biomarkers­—combining biomarkers with the morphology—and that gives you a clear indication about whether you have an immune system that is strong and that will actually fight the cancer or not. This is extremely relevant because, if you have a strong immune system and the immune system can fight, it can have a tremendous positive impact on your disease. Even if the patient has a stage 3 cancer, with a strong immune system, he or she could have very long progression-free survival versus someone with a very weak immune system—demonstrated by a low Immunoscore—who could die within a year because the cancer cannot be managed.

So knowing the health of the immune system becomes a more and more critical element of managing cancer. And that’s the reason why the whole immune therapy paradigm became kind of the theme in science last year, as one of the breakthrough insights about how to check cancer.

PMO How granular does this drill down in terms of informing such things as drug selection? Is it specific enough to inform such decisions as which biologic is going to do the most for this or that cancer or patient and at what dose level?

Mr Heydler Let me step back and talk about 2 fundamental concepts we are dealing with. The conflict arises between a prognostic test and a predictive test.

A prognostic test basically is not related to a specific treatment, whereas a predictive test, also known as a companion diagnostic, is. It tells you which patient group is suited to that specific treatment and it asks you to select the treatment for a certain patient population. In the context of immuno-oncology, I think both prognostic and predictive tests are relevant. First of all, the Immunoscore is a prognostic test, and the prognostic test tells you whether the patient has a strong or a weak immune system. If the patient has a strong immune system, and you have a phase 1 disease, you don’t have to do anything but wait and watch. If the person has a weak immune system, it basically recommends that you provide treatment. But it doesn’t tell you, since it’s prognostic, which treatment to pick; however, if you arrive at the decision to treat, then you have to make a decision about what to use.

For the next step, you would go through a patient stratification with a predictive test, which allows you to select and determine which treatment you want to use that will strengthen the immune system. This means you take the brake off the immune system by taking the camouflage off the cancer—PD-L1 principle—to improve the immune system, whatever the treatment is.

At Definiens, we are also working together with quite a number of biopharma companies on predictive tests to help them stratify patients in the true personalized medicine sense by identifying the best possible treatment. Basically all the major players are embarking on immunotherapy because everybody thinks that’s the way to really move cancer from a really bad disease to a manageable disease.

This is public information. Merck is working on an anti-PD-1 drug. Novartis, Amgen, Roche, Bayer, GSK, and AstraZeneca are all working on this aspect of the disease, whether it’s an anti–CLTA-4 or an anti–PD-1. You can see there’s a massive wave of R&D money in anticancer therapy. This is where Definiens comes in, because you need a full understanding of the phenomic information to come up with a prognostic score. And you need the same phenomic information to come up with a predictive companion diagnostic. This provides dramatic value to the overall immunotherapy story.

PMO What has your exploratory market research into the knowledge level of oncologists and pathologists told you in terms of the clinician’s ability to translate these data into practice and to bring this into practice? Is there a knowledge gap that needs to be addressed, how large is it, and have you looked at a strategy for closing such a gap?

Mr Heydler First of all, we as an industry (and I’m talking collectively about the healthcare industry, including biopharma and diagnostic companies) are in the beginning of that learning curve. I would say in general that when you talk to pathologists or oncologists, everybody understands the importance of the immune system. So I think the generic knowledge is there.

Also, everybody understands that if the immune system is so paramount in fighting cancer, then if you provide the wrong chemotherapy, which basically kills a lot of biological systems, including the immune system, it’s actually counterproductive. That is why I think you see a lot of recurrence in cancer once you’ve done one chemotherapy, because the cancer you’re treating is mutating and is kind of sneaking out—and you have killed the immune system, which can no longer fight the cancer, putting the patient in a bad situation. I think everybody understands that strengthening the immune system is really important.

Now, how do you specifically profile and diagnose a patient? I think this is where the clinician and the pathologist need help, because they can no longer digest all the data that come out of such a test. So they require computer-aided support. That’s why we are focusing on immunotherapy.

At the end of the day, we, as a company, are making the effort to develop a diagnostic test, whether it’s prognostic or predictive, and to perform the test ourselves so that eventually we can create an information package allowing the oncologist to come up with a specific recommendation about a specific patient with a specific cancer.

But as you know, we are in the beginning of this process. The prognostic side is going through clinical trials as we speak. A lot of biopharma companies have certain candidates in their clinical trials on the therapy side. With that comes companion diagnostics, and that’s something where an oncologist should have those tests in his or her hand and be able to prescribe them to get the exact information about the disease that the patient has.

PMO That is why I asked about the Immunoscore. When you bring this to them, how many blank stares are there and how many are saying, “Aha, I can put this into play.”

Mr Heydler I think there’s no question about putting it into play. Everybody we are talking to is really excited that we have the opportunity to change the way we can help patients. We are extremely well aware that we have to go through a process. You might start with out-of-network reimbursement in the beginning. It would have an impact on finances for sure. I think as more and more of those results come out in the trials, we’ll see the guidelines evolve. I don’t think the payers can prevent anyone from communicating the impact, because there’s so much benefit. And there will be pressure from the physicians, there will be pressure from the patients, because as you know, patient groups and advocacy groups are very active these days, and they’re already on this bandwagon.

PMO You need the critical mass of support to get not only utilization but also coverage. Are you then anticipating research that affects the economic burden of cancer with the impact of Immunoscore to be factored into this?

Mr Heydler We are a relatively small and new company. I am sure it might be worthwhile talking to some of the bigger biopharma companies, because I’m pretty sure that they are currently investing heavily to answer exactly that question about the appropriate value proposition. It was a key theme at the 2014 ASCO Conference, which was all about immunotherapy. The large, powerful companies will drive some of the research, and we’re very happy to participate and learn. But I think our focus right now is to really come up with the test and the solution and the information that will help practices.

PMO How do you envision the research process moving forward, getting traction, and having others continuing to test Immunoscore? How do you envision that happening in an ideal fashion to expedite the implementation of what you brought to personalized medicine?

Mr Heydler As you have already mentioned, the Immunoscore is already undergoing a global validation effort in order to confirm its value and be able to expand its use and acceptance. The completion of the Immunoscore validation will be a major milestone. The nice thing about Immunoscore and immuno-oncology is that the core principles are pretty generic. So it’s not as if we have to fully home in and concentrate on only one cancer type, such as breast or prostate or colon. In principle you can actually tackle these small variations very quickly in quite a number of cancer types. Therefore, I think the great promise of that approach is that we could offer a very generic approach going forward. I think this could be picked up massively in the next months or 2 years.

PMO How unique has this been? Are you alone in this field? It sounds that way. Essentially are you first in class, are you still solo in this field?

Mr Heydler The tissue phenomics approach, which complements the genome-based approach, requires a very strong bioinformatics analytical technology to extract all the relevant information from the tissue. That’s a very unique technology differentiator for Definiens. We strongly believe we have proven that through the large number of projects we have. Just to give you a little data point: all the top 20 biopharma companies are customers of ours on the R&D side, and we have done probably more than 2500 projects to show that we can actually extract information from tissue that nobody else can. The reason for that is our proprietary, patented approach, which we call Cognition Network Technology. This approach mimics the way we humans use our cognitive ways of interpreting things we see.

When you look at your phone, all your eyes see is a certain amount of structures and segments. You know the image, but you also know something about this, and you learn in your life this thing you’re looking at is a telephone. You can see and talk to other people, and only when what you see and what you know come together do you understand what you’re looking at. And that fundamental cognitive principle of what we humans use to interpret images is what we have implemented in our technology. As opposed to pattern recognition, imaging has its tools and can show you only similar patterns. With Definiens’ Cognition Network Technology, however, we can actually understand each element we see in the image, and that allows us to use specific interpretations.

This is our unique value. I think so far as the diagnostic side is concerned, we have a very unique differentiator. We are perhaps in a prime position. On the therapy side, as far as immunotherapies are concerned, that’s what biopharma is focusing on and this is what we are using to support them.

PMO What is the timeline for taking this from the researcher’s end and putting it into the practicing oncologist’s hands?

Mr Heydler There are 2 perspectives. On the prognostic side, we believe that we can bring such a test to market through the georeference lab approach, so you don’t have to go through an extensive approval process, such as you do with the FDA, which takes considerably more time. So the timeline could be in the 2- to 2½-year time frame to take samples on the prognostic side of the test. That will happen primarily in my eyes through a georeference lab approach.

On the predictive side, for companion diagnostics, we will have to go through the FDA approval process, which is intensive. I think the predictive side will take a little longer, and that will be driven by the therapy, by the drug. This will take several years. I imagine it will eventually happen in clinical practice.

PMO The incentive to fast track this is clearly here. What you offer is the ability to avoid clinical blind alleys and to get the care going. I wonder how similar this is to the problem of rheumatoid arthritis. There’s now a multibiomarker test that provides a clear, objective assessment tool of disease progression—something impossible prior to its development. And this reminded me of Immunoscore, because prior to that, rheumatologists were, to a large degree, operating in the dark. They hadn’t anything objective to tell them about disease progression.

Mr Heydler I think that’s a pretty good analogy. The fundamental principles you’re describing are absolutely equivalent.

PMO This is really where I think personalized medicine is going. Before we go, do you have any thoughts about how we would define personalized medicine in oncology?

Mr Heydler I think we have a very clear vision in terms of where personalized medicine is going and also where we can help. I think we strongly believe this is where our franchise is. Phenomic information is paramount to really understanding the disposition and the state of a disease in a patient at a given time.

When you think about genomics information—and it gives you a good prognosis, it gives you information on, perhaps, a genetic mutation—you might learn that you have a certain likelihood that you will develop cancer over time. And only in a very few mutations do we have a guarantee, unfortunately, that you will develop cancer. The BRCA1 gene is one of the few cases where we have a higher percentage of the likelihood of developing cancer. In general, however, we don’t know.

Now we can actually combine the genomic information and really look into the phenomic expression to come up with the exact diagnostic footprint or profile of the patient. And once you have that, you know exactly where the cancer is and how many cancers you have at the same time. Current molecular testing gives you only a very vague analysis of your genome, but what you really are interested in is having an exact analysis of the most aggressive cell line(s) in your cancer. This is a function of morphology, and you have to find the answer in the right place. You can’t just put it in a tube and mix it and come up with your molecular test. You have patients who have that specific cancer—which again is phenomic information!

That said, I think phenomics is critical, and once you can combine genomic and phenomic information, you get a complete diagnostic fingerprint of the patient, allowing you to then apply the best possible diagnosis and therapy to that individual patient. As an example, let’s say you have 500,000 prostate cancer cases in a big database and you have the exact genomic and phenomic fingerprint of each of those 500,000 patients. Now a new prostate cancer patient comes to the oncologist. The oncologist can run the new fingerprints, the diagnostic fingerprints, of this patient, then take that fingerprint back to the big database and match it to the 10 most similar cases. What therapy has worked? What didn’t work? What was the average survival time? What are some of the side effects? Suddenly you have a wealth of information that can provide treatment in a most personalized fashion. Now the treating oncologist can make the best possible decision, based not only on genomics, but also on specific phenomic information.

PMO And that is the punch line! What an improvement this is in our prospects for healing. Please accept our thanks for this remarkable walk through the field of phenomics and how you are harnessing it to aid in cancer diagnostic and treatment processes.

Mr Heydler You are most welcome. My thanks to you in turn for enabling us to communicate information of our work to today’s practitioners, that they can begin to reframe their expectations for treating their patients with the precision that only personalized medicine makes possible.

Related Articles