Is Roche going to “knock it out of the park” or “jump the shark”? Everyone I know is excited, even if many of them are sceptical, about the Feb 20th webinar and what we will learn about SBX. Alex Dickinson appears to be driving most of the buzz, having had an early sneak peek and being able to share some of what he learned. I’ll kick off with some of the backstory to Roche’s latest sequencer, an idea of what they might deliver, and some questions I’d like to see answered on Feb 20tgh. I wrote this before having my own early intro while in SF for PMWC – spoiler alert, my own reveal is embargoed till after the Roche webinar on Feb 20th. And you may want to read Keith’sNava’s and Albert’s musings (although I will do that after posting my piece so there may be lots of overlap).

Get ready for a splash at AGBT – I wish I was at the party, I mean conference.

First the backstory to what Roche might have built: Roche do have some info on their Sequencing Platform Technologies website about their semiconductor-based electronic detection nanopore system: they say they will deliver flexibility and versatility (good for cores and clinics), short and “longer” read lengths (note the use of longer and not long), ultra-rapid data generation and analysis (is another Guinness World Record coming soon – can they beat Euan Ashley’s 5h 2m) and all in a yet-to-be-confirmed cost-effective package. How exactly might they have achieved this?

What did Genia and Stratos have: Two key acquisitions were key to Roche’s ambition to build their nanopore sequencing technology: the acquisition of Genia in 2014, which brought in the semiconductor sequencing chip (read what I said then on this Decibio post by Stephane), and the acquisition of  Stratos Genomics in 2020, which brought in the Sequencing By Expansion (SBX) chemistry. How much of this was driven by their relative lack of success after 454 will be the subject of @AGBT bar or hot tub sessions (sorry @Omicsomics I’m not going to be able to join you this time).

A short diversion to 454 and Illumina: For those of you new to NGS (i.e. post 2020) you may not know that Roche acquired 454 Life Sciences in 2007 at the dawn of NGS. 454’s GS20 was the first NGS instrument and it shined – first with bacterial genomes (Margulies et al. 2005) then with James Watson’s genome, and the first Neanderthal sequences. The GS20 (20Mb – read that again, 20 megabases!) was followed by GSFLX (100Mb), Titanium (600Mb) and GS Junior, but Roche closed 454 and withdrew, to a significant extent, from the NGS market. Their last gasps were in the early 2010’s with a hostile bid to acquire Illumina in 2012, a failed partnering with PacBio to develop a clinical diagnostics sequencing product, and failed partnerships with DNAe and IBM to develop semiconductor-based or nanopore sequencers.

Genia: founded in 2009, Genia was a “single-molecule electrical detection sequencer” initially bankrolled with investment from Life Technologies (I am not diverting to SOLiD!). Early descriptions were of a biological nanopore (analpha-hemolysin transmembrane protein suspended in a lipid bilayer) system that integrated analog-to-digital conversion of the signal. Genia published a proof of principle study in 2016, where they demonstrated 264 well CMOS array running their single-molecule nanopore-based sequencing-by-synthesis (SBS) chemistry developed at Columbia University called NanoTag (Kumar 2012).

This incorporated four polyethylene glycol tagged nucleotides, with the tag released by the polymerase for detection by the nanopore. In some sense this approach was somewhat similar to Oxford Nanopore’s discontinued exonuclease nanopore sequencing method (I think this was bankrolled by Illumina, and probably could have been improved further by using technology like Norfolk Nanopore’s Polonopore clustering technology ;-)). Perhaps the last little factoid that might be important in how successful Roche might be, or the troubles they’ll face, is that Genia was called out as possibly infringing on Oxford Nanopore’s patent estate and the Genia chemistry (not Stratos) was the focus of a lawsuit from UCSC.

Roche acquired Genia for $350M with the aim to deliver a ‘sample-in-report-out’ system, which would have been ideal for diagnostics. At the time they gave no timeline for this development but here we are a decade later and still waiting for the launch.

Stratos: There’s some information on Roche’s website but the original Stratos Sequencing-By-Expansion (SBX) chemistry enzymatically converts DNA fragments into 50x larger molecules called Xpandomers by incorporating X-NTPs (modified nucleotides). The X-NTP nucleotides, “seems like one chemist dared another that something couldn’t be done” [Keith Robison], have large chemical loops or tethers attached to the phosphate group and the base, these are incorporated into the growing DNA strand that will be the template for sequencing. After this “amplification” the X-DNA strands are cleaved to stretch out the DNA for sequencing.

An unmodified DNA fragment would move too quickly through the nanopores, but the Xpandomer takes longer enabling the new sequencer to read with high accuracy, rather than amplifying the signal it increases the time to collect signal. Stratos made claims about read quality a long time ago, but we will have to wait to hear what Roche have been able to deliver.

So where is Roche and what can SBX do today: based on the details released in the webinar announcement we know that the new instrument is still a single-molecule nanopore sequencer, and that it generates 500Mb per second using an 8 million pore array. This will get Roche to >15B >300bp reads in 4 hours (no confirmation of read length and no leak on quality just yet) – which equates to around the 8Tbp that Illumina’s NovaSeq X Plus and the 6-8 Tb from Ultima’s UG100, both in a 24 hour run. The only negatives that leap out to me are the long-reads and direct methylation we’ve come to assume nanopores would give us, but I can live without that for most applications.

If these specs hold true (see table below) in customers hands, then Illumina et al have cause to be worried. With a benchtop footprint when compared to either of the NGS behemoths, this box will likely have lower capital requirements making purchase easier for everyone – another go at democratising sequence is on the horizon.

And with these specs’ genomes (in the NICU and in Oncology), and not just panels, are going to be possible. As a high-throughput system then single-cell RNA, spatial, and proteomics, will all be possible. And as a high-enough accuracy instrument maybe the “killer app” for Roche will be end-to-end clinical diagnostic assays for the SBX sequencing platform. Unlike Illumina, Element, ONT, or PacBio, Roche is a diagnostics company, and the value is in the assay (i.e. sequences) rather than boxes (i.e. sequencers). Of course, this means the new sequencer needs to be accurate in both the raw sequencing (Xpandomers need an engineered polymerase with its own biases) and in the library prep. So the molecular biology of Expandomers needs to be understood better (InDels, substitutions and other errors) before we can say whether there could be issues that still need to be resolved.

But the specs seem to fit what the high end NGS diagnostics market is doing today with an instrument that a lab could easily do 2-3 runs per day – Roche are knock, knock, knocking on FMI, Natera, Caris, Tempus, Personalis, Myriad, et al’s doors.

Some questions I’d like to see answers to:

  • What will the CapEx and running costs be?
  • What quality will they achieve? Whilst I think a lot could be done in a clinical system that had Q20 base accuracy but high sensitivity and specificity to call clinically relevant variants (e.g EGFRm like their COBAS assay), I am acutely aware that quality has gotten much better with the work of Ombniome (PabCio’s ONSO), Ultima’s ppmSeq, and Element’s AVITI. Will users accept a lower quality system?
  • Can we do methylation directly or will we be stuck with bisulfite or enzymatic conversion?
  • Will the system be launched with clinical data from the likes of FMI?
  • How did the early access sites, presumably Broad and Sanger, keep this quiet (although they managed to keep Ultima in stealth for years)?

I will finish up with a quote I have used myself and have heard many times over the last decade: “Roche is the company where sequencing technologies go to die!” I am hoping that we’ve all been wrong, and Roche used their earlier experiences to design an instrument they believe could deliver diagnostics to users across the world, which may well be in our hands in the not-too-distant future. Exciting times indeed!