Credit: Anand Purohit / Getty Images / Moment
For the past year and a half, Roche Diagnostics has been drumming up interest around AXELIOS, its return to next-generation sequencing. The platform officially launched in late June and came out of the gate with an attention-grabbing result: a new world record for clinical sequencing of an entire human genome. But in NGS, better technology—at least on paper—does not always win. Viewed alongside the recent FDA approval of Roche’s blood-based Alzheimer’s biomarker test, AXELIOS looks less like an isolated return to sequencing and more like one piece of a broader diagnostics strategy.
Palani Kumaresan, PhD, Global Head of Roche Diagnostic Solutions [Roche Diagnostic Solutions]For Palani Kumaresan, PhD, global head of Roche Diagnostics Solutions, that strategy spans the disease continuum, from risk assessment and screening through definitive diagnosis and recurrence monitoring. Roche is even trying to automate mass spectrometry, a technically demanding technology traditionally concentrated in specialized laboratories, and make it accessible to more routine clinical labs.
An Alzheimer’s blood test, a potentially disruptive sequencer, and a more user-friendly mass spectrometer may seem like unrelated bets. Together, they point toward a common goal: making sophisticated diagnostics practical enough for everyday medicine. Roche also has an advantage in getting there, as its Cobas instruments already have a massive footprint in clinical labs around the world.
So will AXELIOS displace Illumina and its enormous installed base? After investing heavily in AXELIOS and following its earlier experience with 454 and pyrosequencing, will Roche turn impressive technical performance into a meaningful challenge to Illumina’s dominance? For now, it is simply too early to tell.
This interview has been edited for length and clarity.
IPM: As disease-modifying therapies for Alzheimer’s are becoming available, how does that change the need for diagnostics? What could a blood-based test like pTau217 make possible that existing Alzheimer’s diagnostics cannot?
Kumaresan: Now, two Alzheimer’s therapies are in the market from other pharmaceutical companies, and we have one in the late stage that we are very excited about. As these disease-modifying therapies are coming to market, there is clearly a very important focus on diagnostics and definitive diagnosis of Alzheimer’s. The standard of care for diagnosing Alzheimer’s today is you will typically do a PET-CT scan, which is quite prevalent in the U.S. If you look at European markets or other places, you will do a cerebrospinal fluid test. You will do a spinal tap, you will do a lumbar puncture, and then you will essentially test for certain biomarkers. This is quite invasive. PET-CTs… are expensive too. Typically you would not recommend these diagnostic tests to an individual unless they’re showing very clear symptoms of dementia… But by then, a lot of these diseases have progressed significantly.
The question becomes: As these therapies are coming, can you have a blood-based test that’s a simple test like a phlebotomy? Can you do a simple blood test that can give you, with high sensitivity and high specificity, confidence that someone is actually having Alzheimer’s disease so that you can then do the subsequent workup?
There have been some tests in development, but these are either launched only in certain geographies like the U.S., offered through a certain CLIA lab…or they run on instruments that are not broadly available in the market. We have the world’s largest installed base of our immunochemistry platforms, Elecsys. We have thousands of these platforms.
We recently got the FDA approval… Two months back, we got the approval in European countries, what we call the CE-IVDR approval. It’s a simple blood test. It’s called pTau217, a phosphorylated version of tau, and it’s really looking at the amyloid pathology. It’s a very simple test, which is approved for running in individuals who are showing early signs and symptoms of dementia. And it can run in both the primary-care and secondary-care settings. That’s very important because most of these individuals will first show up in a primary-care setting… In a primary-care setting, you need to have very high specificity… versus in a secondary-care setting, you need to have very high sensitivity. You want to make sure that you don’t miss a person who has been referred to a secondary-care setting.
With this antibody that we have and the test we have developed, we have shown in a very broad-based trial across multiple sites in multiple countries that it has very high sensitivity as well as specificity in this patient group in both primary- and secondary-care settings. Like any other test… we’ll always have a gray zone where you cannot quite make out whether they have or do not have Alzheimer’s. You want that gray zone to be as small as possible… This test, besides having high sensitivity and specificity, also has a much smaller gray zone, and it runs on a platform that is very widely available.
We have several other biomarkers in the works because ultimately the thinking is, “If you have the right biomarkers, could you be getting it tested if you’re above a certain age…even if you don’t have any signs and symptoms of dementia? Could it just become part of our routine checkup?” That would be ideally the place to go. We’re not there yet, but that’s something that we continue to work towards.
IPM: How do you account for population differences when developing and validating biomarkers like pTau217, and does that work continue after a test reaches the market?
Kumaresan: We want to have multiple sites across the world that are more representative of the population to which we will be selling these tests… We cannot do all of it, so we do as best as we can. The journey continues post-launch. Once we have launched the product, we work with different investigators who want to study these tests in their own population, and that can sometimes even result in different cutoffs, which can be population-specific as well.
Even with machine learning, artificial intelligence, and other technologies, we are doing it in a way that enables people to look at the population they serve. For example, if a lab serves a particular population in, say, India, and you can see how the cutoffs for some other tests differ for that population compared to the rest of the world, you can get a lot of insight into how cutoffs for new tests might behave in your population as well.
IPM: Roche has a complicated history in sequencing, including 454. What makes AXELIOS different from Roche’s previous efforts, and what do you think gives it an advantage over sequencing-by-synthesis platforms?
Kumaresan: Roche has a long history in sequencing. We approach the field with humility because we have had our successes and failures. When we were going from Sanger sequencing to sequencing by synthesis (SBS), there were two companies, essentially. There was Illumina, and there was 454. Roche acquired 454 at that time. Of course, 454 was not successful. But we learned a lot from that.
Now, AXELIOS is really two important acquisitions coming together. One is the Genia Technologies acquisition in 2014. From that comes our sensor module, the sequencing chip that forms the foundation, which is this 8-million-well chip on which the sequencing is performed as single-molecule sequencing.
Then there’s the Stratos Genomics acquisition that we made in 2020, which is this novel chemistry. You are literally translating the DNA into a surrogate molecule, which is much bigger and has greater spacing. When you thread it through the nanopores, you get better resolution… The speed comes from how fast the DNA can go through these nanopores, but that is also an issue because you lose resolution. That’s what we have solved with sequencing-by-expansion (SBX) chemistry… You get higher resolution. You get the best of both worlds.
The launch happened towards the end of June. Because it’s such a different technology and chemistry versus what the field generally knows today, we clearly understood that we had to do a lot of homework before we launched the technology… That’s why we collaborated with Broad Clinical Labs (BCL) as well as Boston Children’s Hospital to show the power of this technology. The fastest human genome previously was sequenced in five hours and 20 minutes at Stanford, but the quality of the sequence… how accurate your base calls are, was much lower… This particular research study, working with BCL and Boston Children’s Hospital, was able to show that in under four hours, you can have really high-quality base calls. And the sequencing part of it was only 20 minutes.
That was just one demonstration of the technology. What our customers and people who have been early adopters like is that enormous throughput—the amount of sequence data that you can get in such a short period of time.
You also need to have real-time data analysis. We have brought that in, working closely with Nvidia…to drive the acceleration of the initial base calls, but also the post-primary analysis…on GPUs, on the hardware, in real time, so that you can really get the power of this ultra-high throughput.
Also, you can start and stop the sequencing whenever you want… With SBS, the data only comes out after you have done all the steps for all of those [DNA fragments] in parallel. You have to wait for 24 to 48 hours before you can start processing the data. But here, you have already converted the DNA molecule into the surrogate molecule first, and then you start sequencing. Your rich data starts coming through pretty quickly, with more flexibility.
In a clinical setting, this becomes very important because you don’t have to wait for a lot of samples to come through before you start a big run. People do this batching for cost reasons. But now, if you can run it in a more flexible way, you don’t have to go through this batching. That means, for patients, you can run the workflows much faster and get the results out much faster.
There’s interest in the market, both from a research and clinical standpoint, based on whether you want answers fast or you’re trying to do ultra-high-throughput sequencing like single-cell work or spatial genomics work, which are very sequencing-hungry… We are seeing a lot of interest from both the academic side as well as the clinical side.
IPM: As precision medicine becomes increasingly data-intensive, how much are sequencing costs and throughput still limiting what researchers can do—and how does AXELIOS change that equation?
Kumaresan: SBS has served us very well for the past 20 years, but it is plateauing out. The other thing that excites our customers is not only seeing what we can do today with our technology, but they’re also seeing how much more this technology can improve, in orders of magnitude, in terms of the throughput and the speed, which will then allow doing these multi-omics applications at cost points that will become much more accessible…
That’s what we also see with a lot of academic groups that we are talking to. They feel like when we announced one million reads for $0.06, which is essentially a significant improvement versus what you can achieve with the largest-throughput instruments today that are available in the market, which got researchers thinking in different ways about what they can do. Because today, even what they want to explore is constrained by how much sequencing they have in their hands and the budget they have in their hands.
IPM: Roche is trying to bring mass spectrometry (mass spec) out of specialized, high-complexity labs. How are you making the technology more accessible, and where do you see it fitting into precision medicine in the future?
Kumaresan: This was a challenge that we took on a little over 10 years back… Mass spec has always been in high-complexity labs. You have to have typically individuals with PhDs running these instruments. Can you bring it to a medium-complexity lab in a highly automated way? And that’s the challenge we took on along with our partners, Hitachi… We worked with them to build this instrument that we launched a couple of years back in different markets. We actually launched it in the U.S. with certain analytes and continue to build that menu on this platform.
At the end of the day, the bottom line was to really democratize mass spec, to bring this technology to more people, so when they go in to get their blood tested for any of these different analytes, they have access to the best technology for those analytes. In the context of precision medicine, mass spec can also be useful if we are looking at post-translational modifications… That’s something that we will explore in the future.
IPM: How do you close that gap and make advanced diagnostic technologies more accessible to patients around the world?
Kumaresan: How can you make it very modular, make it simpler to use, so that you can truly democratize these instruments and the use of these instruments around the world? Diagnostics really influences over 70% of decisions that you take in the healthcare system. But the amount of money, in terms of reimbursement and so on, that goes into diagnostics is only three percent to four percent of healthcare costs, typically. There is a big mismatch in how much diagnostics plays a role in managing disease and how much it actually gets paid.
Being both a therapeutic and diagnostic company, we strongly believe that we need to make sure that diagnostics get out there. There are multiple ways to get it out there… Near-the-patient, point-of-care diagnostics is very important, and that’s something that we focus on too.