Roche’s AXELIOS 1 Brings a Reusable Sensor to Next-Generation Sequencing
Roche is taking AXELIOS 1, its single-molecule next-generation sequencing platform, into Asia Pacific following a global launch in June 2026 and early adoption in South Korea. The company will hold AXELIOS Days in Sydney on 9 and 10 November and Seoul on 12 and 13 November, bringing researchers and laboratory leaders together for scientific sessions and live demonstrations. The platform is the first to use Roche’s proprietary sequencing by expansion technology, and the specification that matters most is not the throughput figure but the word reusable attached to its sensor.
How SBX actually differs
Most sequencing methods read a signal generated at the molecule itself. SBX does something structurally different: it encodes the sequence of a DNA or RNA molecule into a surrogate polymer called an Xpandomer, roughly fifty times longer than the original fragment, in which the sequence information sits in high signal-to-noise reporters.
Length is the point. A longer molecule threaded through a nanopore produces a signal easier to distinguish base by base, because each base’s contribution occupies more physical distance and arrives with less overlap from its neighbours. That is the accuracy argument, and it is a genuinely different route to accuracy than the incremental chemistry improvements that dominate the field today.
The instrument measures Xpandomers using a complementary metal-oxide-semiconductor sensor containing millions of nanopores. Roche describes the sensor as reusable, which is the detail separating this architecture from flow-cell platforms where every run consumes the consumable.
The numbers and what they mean
Roche publishes a run time of approximately fifteen minutes to four hours, at least 1.8 terabases in a four-hour run, reads of up to about 1,500 base pairs under appropriate sample and library preparation conditions, and the potential for whole-genome results within the same day.
Each number answers a different objection. Fifteen minutes addresses the laboratory that needs an answer now. Four hours at 1.8 terabases addresses the laboratory that needs volume. The 1,500 base pair read length addresses work with structural variation or repetitive regions where short reads fail. And same-day whole genome addresses the application that has been the field’s ambition for a decade.
The claim with the most consequence is the last, and it rewards precision about what Roche has said. The release describes same-day whole-genome results as a capability the platform can produce, and cites a 2025 New England Journal of Medicine paper on same-day genome sequencing in critical care. It does not claim a turnaround time for a specific workflow, and it does not quantify accuracy at all.
The competitive position
The platform enters a market with established positions. Short-read next-generation sequencing is dominated by instruments with enormous installed bases and validated workflows, and long-read sequencing has carved out structural variation and repetitive regions. AXELIOS 1 claims a middle position: reads long enough for structural work at 1,500 base pairs, throughput high enough for population work at 1.8 terabases, and run times short enough for time-critical applications.
A middle position is commercially difficult because it competes on two fronts at once. A laboratory doing routine short-read work sees a platform whose throughput it may not need. A laboratory doing long-read work sees reads shorter than the platforms it already trusts for that job. The case for AXELIOS 1 has to rest on flexibility rather than on winning either category, which is why the small-batch economics matter as much as they do.
The reusable sensor is the commercial story
The reusable CMOS sensor deserves more attention than the release gives it, because it determines whether this platform changes laboratory economics rather than only laboratory capability.
Every sequencing platform has a consumable. In flow-cell architectures the consumable is the flow cell, and its cost per run sets a floor beneath which a run cannot profitably go. That floor shapes which experiments happen: it is why laboratories batch samples until they have enough to justify a run, and why small, time-sensitive projects cost a lot per sample even when the science is simple.
A reusable sensor changes that arithmetic if it holds at scale. The release says the platform suits smaller, faster batches or larger projects without changing systems or workflows, and that flexibility means something commercially only if the cost of a small run falls. Roche publishes no cost per run, no sensor lifetime and no replacement interval, so the claim remains architectural rather than demonstrated.
What 1,500 base pairs buys
The 1,500 base pair read length deserves a note because it sits in an awkward band. It is far longer than the 150 to 300 base pairs typical of short-read instruments, enough to span many structural variants and resolve repeats that defeat short reads. It is far shorter than the tens of thousands of base pairs long-read platforms produce, which is what makes those instruments useful for assembling across large repeat regions and for phasing.
Roche does not claim the long-read territory, and the release describes 1,500 base pairs as achievable under appropriate sample and library preparation conditions rather than as routine. What the band buys is speed and throughput at a read length where many biologically interesting variants become visible, and that is a defensible position provided the accuracy holds under independent testing.
APAC and the diversity argument
The regional framing rests on a market figure: MarketsandMarkets projects Asia Pacific next-generation sequencing to more than double from 3.4 billion dollars in 2025 to 7.38 billion by 2030, a forecast the release cites.
The more interesting argument is one the release gestures at without developing. It describes researchers pursuing studies to understand complex diseases and the genetic diversity of regional populations. That matters because reference genomes and variant databases remain overwhelmingly of European ancestry, and a platform deployed across Asia Pacific institutions is a route to correcting the imbalance, provided the data flows into shared resources rather than staying inside individual laboratories.
Roche names two regional customers. The release quotes Macrogen chief executive Changhoon Kim on the Korean service provider taking on projects previously difficult to pursue. It quotes Australian Genome Research Facility chief scientific officer Cath Moore on speed, scalability, flexibility and read length. Both are service organisations, which means their adoption signals throughput confidence rather than one laboratory’s experiment.
The research-use boundary
Every claim here sits behind a sentence appearing twice: for research use only, not for use in diagnostic procedures.
That boundary does more work than a standard disclaimer usually does, because same-day whole-genome sequencing is precisely the capability clinical genomics has wanted for critical care, and the release cites a critical care paper to make the point. A platform delivering a genome within a shift would change how clinicians practise rare disease and neonatal intensive care. It is also a platform not validated for diagnostic use, and the gap between those statements is where regulatory approval, analytical validation and clinical evidence all live.
Nothing in the release crosses that line. The quotations come from research and service organisations, the applications described are research applications, and the RUO statement is unambiguous. A reader should still notice how close the marketing sits to the clinical use case.

The analysis stack
The XOOS analysis suite is a small detail with a large implication. A sequencing platform is only as useful as the pipeline turning its output into variant calls, and laboratories do not lightly adopt a new analysis stack. Offering a free open-source suite removes the software cost objection at the point of evaluation, and describing it as open source without naming a licence is the one place the release leaves an obvious question open.
If the suite carries a permissive licence, a laboratory can inspect and modify the variant-calling logic, which matters when a research question needs a non-standard pipeline. If it is open source in name only, the benefit reduces to a free download.
What is missing
No instrument price, no cost per run, no sensor lifetime or replacement interval, and no consumable pricing for the library preparation kits. No accuracy specification appears anywhere: no read quality score, no error rate, no consensus accuracy figure. The 1.8 terabase throughput carries no qualifier about read length or quality at that output. The release states no number of early-access sites, describes the South Korean adoption without a customer count, and names no licence for the open-source XOOS analysis suite.
One more gap deserves naming. The release does not say what sample types the platform accepts or what input quantity a library preparation requires. For a laboratory working with liquid biopsy samples or single cells, input requirement is often the constraint deciding whether a platform is usable at all, and the tested applications list includes single-cell RNA without stating what that required.
What to watch
Whether Roche publishes a cost per run and a sensor lifetime, because those two numbers decide whether the small-batch flexibility is real. Whether accuracy specifications appear in a peer-reviewed comparison against established next-generation sequencing platforms, since the Xpandomer approach is novel and novelty demands benchmarking. And, hether the regional deployments generate sequence data entering shared variant databases, which is the difference between selling instruments in Asia Pacific and addressing the diversity problem the release raises. And whether the diagnostic pathway opens, because the RUO boundary is the entire distance between a research platform and a clinical one in practice.

