Home PublicationsData Innovators 5 Q’s for Gabe Otte, CEO of Freenome

5 Q’s for Gabe Otte, CEO of Freenome

by Eline Chivot
5 Q’s for Gabe Otte, CEO of Freenome

The Center for Data Innovation spoke with Gabe Otte, co-founder and chief executive officer of Freenome, a U.S. biotechnology company that uses blood-based multiomic tests for early cancer detection. Otte discussed how new methods to digitize and examine a much broader range of molecules in the blood can support the analysis of patterns associated with specific types of cancer.

Eline Chivot: Freenome operates a “multiomics platform.” Can you explain what it is, and how it can help in early detection of cancer?

Gabe Otte: From a broad perspective, early detection of cancer is the key to curing cancer. For most cancer types, the difference between detecting cancers early versus detecting them late is about chances of survival—over 90 percent versus 10 percent. When the company was founded in 2014, we settled to do this differently than other companies. We believed there was no simple solution to cancer screening, and that there wasn’t going to be just one or two biomarkers allowing us to detect cancer early. We thought that if it was that easy, this would have been a solved problem already. We decided to build the multiomics platform to look at as many biomarkers as possible in the blood. The idea was to create data from a single tube of blood across the multiple analytes or signatures that you can look at, whether it’s DNA, RNA, proteins, etc. Our bet was that if we took a really good look at as many signatures in the blood as possible, we would be able to create the most accurate and effective early-detection test possible from blood.

The platform is a combination of what we call the wet lab and the dry lab. The wet lab technologies we have developed aid in preserving multiple signatures within one blood collection tube. Most tubes cannot do that. We then apply all sorts of automation through a set of entirely new analyses.

So once you have the data coming out, you have to combine multiple signatures. If you’re using DNA and RNA and try to combine them, the complexity lies in what you report to the patient: The DNA results may show that a patient has cancer, while the RNA shows that he or she does not. What do you then report? It makes sense to look at multiple signatures, but combining them so that they’re actually additive as opposed to contradictory is a challenge in and of itself. This required a lot of work from our machine learning teams and took us two and a half years to solve. 

Chivot: Freenome isn’t the only company developing early stage cancer screening tests, but you are one of the few using machine learning. Can you explain the merits of using the technology to detect early stage cancer? 

Otte: The way we use machine learning is probably different than the way others do. The focus of our machine learning is on what we call additivity—i.e., as described above, how you add the different analytes together.

Machine learning is not a kind of magic black box, it’s quite simply a computer doing statistics. Humans are very good at statistics especially when they can visualize the data. For instance, if one is looking at a single analyte, one can color the cancer dots and the healthy dots, identify the line that divides those two, and come up with a test when it’s segmentable.

For cancer screening though, the problem is so complicated that a single analyte wouldn’t be enough to distinguish the separation between cancer and healthy dots. We then have to use multiple analytes, meaning we have to look at multiple analytes at the same time. That’s drawing that same line of separation but in a multidimensional space—and human beings aren’t good at visualizing various dimensions.

That’s where machine learning is particularly powerful. As we feed a system with more and more data, it gets more sophisticated and can separate patients with cancer from healthy patients in a multidimensional space. Because of the way we have built the tests, the tests improve their performance over time—which has traditionally never happened. In fact, what has happened is the opposite. A test called PSA—prostate specific allergen—which has been used for prostate cancer detection for the last 30 years recently fell out of recommendations for clinicians to use because while it had launched with decent performance in terms of accuracy (50 to 60 percent), today, it is down to 20 to 30 percent. That has happened with other different tests, including mammography for breast cancer detection.

Machine learning-based diagnostics is a future where the more data you get, the more accurate a diagnosis becomes, and the more lives you can save. This is something I’m really looking forward to releasing. It’s never happened with molecular diagnostics, no one has done it with a blood test to date but it is something that we are working on very closely with the Food and Drug Administration (FDA).

Chivot: Getting this technology to market requires expensive and long clinical trials. But do you think this is evolving towards broader accessibility?

Otte: We’ve recently announced that we launched a 14,000 patient clinical trial and will be submitting it to the FDA for validation. That’s the largest trial ever done in colorectal cancer screening. It does take a lot of resources, and a lot of time to collect patients, samples, etc. It is a real challenge to make sure your test is safe and working as expected. That’s all traditionally been done through expensive clinical trials. The openness of regulatory bodies like the FDA allows us to progress with such breakthrough initiatives. Indeed, regulators today are ready to recognize that we can’t afford to have tests that decrease in performance, and that we need to find better solutions for patient treatment.

The idea is to get to a point where we can use real world data to have new tests approved, instead of having to run more clinical trials that are very expensive and laborious.

Chivot: Ultimately, isn’t Freenome also about advancing personalized medicine? What are some of the biggest obstacles for this?

Otte: I very much believe that we’re furthering personalized medicine. In fact, that’s precisely our strategy. In our space, other companies working on cancer screening are trying to do these one-size-fits-all multi cancer tests, which comes with many drawbacks because these tests are not inherently personalized, and each cancer type is treated so differently. That’s one of the reasons why Freenome decided to be a multicancer company with multiple tests. We want to do right by each cancer type, so that we can focus on the patient at hand.

Somebody who has colorectal cancer is dealing with very different challenges from somebody who has pancreatic cancer. The tests and the types of therapies therefore have to be very different. We’re just at the beginning stages of personalized medicine for early cancer treatment, because today, most cancer treatments happen at the late stage of the disease, including because 80 percent of all cancers are themselves detected at a late stage. Most cancer drugs, whether it’s immunotherapies or chemotherapies, are currently designed for late stage treatments.

As companies like Freenome progress, they will shift the entire paradigm of how we treat cancer from late stage to early stage. As we shift towards early stage detection and treatment, we’re going to need all sorts of new tools, including new personalized medicine diagnostics tools—as indeed most tools today are only developed to address late stage patients. Working in early stage is an entirely new field that’s being created by early detection programs. As early detection includes screening and early intervention, we work hand in hand with pharmaceutical companies to make sure that we can ensure the follow up of patients after detection, and that they are provided the right treatment that will work best for them.

Patients are open to innovations like ours and personalized medicine, as they always want the best tests and treatments. But at least in the United States, some companies in this space will tell you that high specificity is what matters when it comes to screening (and it does matter), meaning you don’t want false positives, as they lead to undue financial burden on the healthcare system. But they sometimes overlook the implications of false negatives, and prioritize high specificity over sensitivity. From a patient’s perspective, false positives aren’t as bad, generally speaking, as they ultimately find out they are healthy, while false negatives mean although the tests are negative, you have cancer but think that you are healthy, which is far worse.

Thus, it’s really important to focus on the right numbers for the right applications. Some cancer types will require high specificity, others will need to optimize for sensitivity. It’s all about understanding the context.

At Freenome, at least for colorectal cancer screening, we believe that it’s important to have both high specificity and sensitivity tests. In our latest prospective trial results in January, our tests demonstrated a 94 percent sensitivity and a 94 percent specificity. We didn’t sacrifice sensitivity nor did we sacrifice specificity, we didn’t settle for a trade-off, we just did better. We want our products to be commercially successful, yes, but you can’t forget about the patient and who the patient is. We want to make sure we do what it takes to make the company successful, but that also entails thinking about the patient.

The onus is on us collectively as a people to make sure we’re doing right by the patients as much as possible when it comes to medical care, and that involves collaboration with policymakers, industry leaders, and regulatory bodies. One of the central tenants of Freenome’s culture is empathy, because similarly, a multiomics platform does not get created by a single person. It is the result of the work of different teams of people—molecular biology, computational biology, machine learning, software engineering, clinical laboratory science, regulatory reimbursement, business development, legal, etc.—all coming together, and they don’t even speak the same language, half the time. We created this culture of empathy so that we have people that are very good at what they do, but also humble enough to learn from each other on how to move forward into the future. Pushing for a future that improves healthcare for all is going to take that collective empathy of all of us, including your readers, including policymakers, working with industry leaders like ourselves. I look forward to working with everyone in that capacity, in this spirit of humility and understanding each other’s perspective to achieve this mission.

Chivot: How could Freenome and its multiomics platform further revolutionize future disease management? What does the next generation for future applications look like?

Otte: Our multiomics platform allows us to look at more signatures in the blood than anyone else, and we’re not going to stop at colorectal cancer. Freenome was founded to cure all cancers and diseases. Part of our platform is about understanding and characterizing the immune system and how it is responding to different diseases. Your immune system really is the ultimate diagnostic: It is your body’s way of figuring out what is going wrong. Meaning that the better we understand that immune system, the better we can diagnose all sorts of diseases, not just cancer.

If you look back 10, 15 years, most clinicians would say the concept of using your immune system to go after cancer cells is ludicrous. Yet today, we do it routinely, we program immune cells to go after tumors. Today, it is a ludicrous concept to use the immune system as a diagnostic to detect all sorts of diseases accurately. But ten years from now, diagnostics companies will be able to detect all sorts of diseases early and when it’s treatable, using the immune system and other signatures in the blood. Therapeutics companies will program the immune cells to go after all sorts of diseases. This is really about utilizing all available signatures and tools available to not only to detect a disease well but also to treat the disease as early as possible. I think the future lies in understanding these signatures well enough, so that we can diagnose and treat all sorts of diseases.

You may also like

Show Buttons
Hide Buttons