Roche has entered the conversation…

It’s been an exciting couple of years for the sequencing market. Unless you’re a sequencing company, especially one named “Illumina”, and then maybe it’s been more terrifying. But for sequencing customers and those of us in the cheap seats, it’s been a fun ride compared to the previous five years or so when Illumina was skating by without any real competition. And now Roche is poised to add to the excitement/terror. They’ve announced a webinar on Feb 20th, just ahead of AGBT (at which they’re a Silver sponsor) with the prosaic title “Introducing Sequencing By Expansion (SBX)”. While the title may not stir the blood, the description has sparked quite a bit of interest in the genomics community. It’s been fun to see all the speculation about what Roche really has and what it means for the market. I can clear up some of the misconceptions, but we’re going to have to wait until the 20th for most of the details to go public.

Whenever a company gives a presentation about future products, there’s a high risk of getting a whole lot of “marketing fluff” (I’m allowed to say this as I’ve spent a lot of time in Marketing). This is often done by startups to help spur investor interest. But Roche is not a startup company and as far as I know they’re not raising funds. I’ve seen a dry run and this is a science and data heavy talk. It will be presented by two Roche heavyweights: Mark Kokoris (Stratos CEO) and John Mannion (VP Computational Sciences). So be ready with some tough questions and let’s see what they’re allowed to say!

I’ll save the history of Roche’s foray into sequencing for another post, but suffice it to say that they’ve long been derided as the company “where sequencing technologies go to die” (a phrase so common it was likely thought up independently by many). So, given this reputation, why are people so excited? And should they be? Here’s how Roche has described their pending SBX sequencer: “Highly accurate whole genome sequencing performance with over 15 billion reads in four hours of sequencing.” In addition to WGS, they also mention RNA and single cell RNA samples with “longer reads than a traditional short read platform”. There’s a lot to unpack there, so I’ll break it down below.

Technology

Roche’s sequencer has its origins in the acquisition of two companies, Genia in 2014 and Stratos in 2020. Genia came with a protein-based nanopore platform that has scaled to 8M sensors and a detection chemistry that, frankly, never panned out. (That said, Nava Whiteford has uncovered some Chinese companies that appear to be copying the original Genia playbook. I wish them luck.) Roche complemented the Genia nanopores with the Stratos “sequencing by expansion” (SBX) chemistry. It converts the DNA to be sequenced into molecules that are ~50X larger than the original DNA molecule. This is done via a polymerase which incorporates highly modified nucleotides with large sidechain loops. Once the molecule is created, the backbone is snipped and the loops expand. This expansion, along with some pore transmission control elements, creates a molecule that’s much easier for the nanopore system to read. This is where Roche’s claim of “highly accurate” comes from.

Read length

When most people hear “nanopore” they naturally think of Oxford Nanopore, which reads native DNA molecules and can have incredibly long read lengths. But Roche isn’t reading native DNA. They’re reading the “expandomer” molecules described above. Roche hasn’t officially said what the read length is, but there are some clues. First, they described them as “longer reads than a short read platform”. Since they’re positioning against short read platforms, which tend to be in the 150 to 300b region, that’s the ballpark we should be thinking about. The second clue is that Stratos used to talk about 500b molecules. Those were early days and I’m not sure if they ever actually achieved that length, but it’s still in the “a bit longer than short read” territory. 

Speed

A four hour runtime is fast, but a speed that has been achieved before. Thermo’s “Do they still sell those!?” Ion Torrent-based sequencers have been capable of short runs since they launched in 2010. And Oxford Nanopore doesn’t really have a set runtime, so they can be as short as you want - it just limits how much data you get. But getting 15 billion reads in 4 hours is unheard of. This could allow for really high throughputs (see below), but it could also be beneficial for clinical applications where “time to answer” is critical. It could also fit better into a normal 8 hour work shift. That said, in a “sequencing factory” setup, Ultima Genomics’ “walk away” setup might be more advantageous. 

Throughput

This one is a little harder to estimate as Roche hasn’t said how long the reads are, just that they are “longer reads than a traditional short read platform”. Since Illumina reads are traditionally 2x150, it’s pretty safe to assume something in the range of 300b for Roche. Maybe longer, but let’s stick with 300b for now. 15B 300b reads = 4.5Tb, or 1.125 Tb/hr. That’s about 3.5x the throughput of a maxed out Illumina NovaSeq X running 25B flow cells or a maxed out Ultima Genomics UG100. If instead we go with Roche’s statement of 500Mb/s, that comes out to 1.8Tb/hr, or ~5X the throughput of Illumina and Ultima. 

Read Quality

Roche hasn’t provided many clues here. Pretty much all they’ve said is “high quality”. Guesses over on the ASeq Discord seem to have ranged from Q20 to Q50 (although I think the Q50 “guesses” were more a statement of what would be required for clinical applications like MRD). Personally, I don’t think a Q20 sequencer would do well. It -could- work as a counting machine for single cell RNA, but that feels a little too specialized. I think general sequencers need to be at least Q30 and then have a path to Q40/Q50 consensus reads.

Cost 

Roche really hasn’t given any clues here in terms of the instrument cost or sample cost. Because this is a non-optical system there are reasons to be optimistic that it can be relatively low cost, but how low can it be and how low does it need to be? It’s unclear what, if anything, Roche will reveal about pricing during the webinar. Which relates to the next topic…

Launch Timing

Same lack of clues. At the 2024 JPM Healthcare conference they listed nanopore sequencing as a “2025+” launch. Given that statement and the nature and timing of the webinar, a 2025 doesn’t seem too likely. It will be interesting to see if they give an updated timeline at the talk.

Market

In addition to the timing of the launch, there’s also the question of what market they will be launching into. There’s been quite a bit of speculation that the Roche nanopore sequencer will be launched as a clinical solution, possibly only for use with Roche assays. While I didn’t think this was true, it’s not completely out there. Roche is, after all, a clinical company. But there are strong clues that this is not the case - the talk description mentions whole genome sequencing, RNA, and single cell RNA. That does not sound like a platform being readied for a clinical launch. But I wanted to be sure, so I asked. Roche confirmed that the intention is to launch this as an RUO (research use only) platform for general sequencing. So start thinking up your favorite use cases for this sequencing beast.