Roche Ripple Predictions

In the prior piece, I covered the technical details unveiled by Roche for their SBX technology, but generally tried to avoid predicting its effects on the marketplace.  Here I put on the pundit’s hat.  The TL;DR is this is a major new sequencing platform and if you’re at one the competitors you have about a year before it fully hits the market - though in reality the action has already started as Roche starts grabbing hearts-and-minds.  What can we anticipate about the effect on each of the current players? As noted in the prior piece, some key aspects - in particular purchase price and run cost - aren’t being disclosed by Roche and complicate prognostication.  

If you're at a competitor, the glass-half-full view is that this technology isn't yet a platform and your customers can't buy it yet.  But they can be distracted by it - or worse, start putting off capital purchases and there are early access sites at Broad Institute and Hartwig Medical Foundation generating results.  So you have both have time and don't have time to determine a strategy to counter SBX.

Illumina

For Illumina, this is a serious moment.  One way to partition the sequencing market is those who insist on Illumina (“nobody got fired for choosing Illumina”) and those willing to entertain one of the other players.  An easy way to herd back into the “only genuine Illumina” camp would be to suggest that the competitors are all on shaky financial ground; who wants to be left high-and-dry if such a company exits the marketplace?  Or less dramatically, which competitor has the global reach and logistical prowess of Illumina?  

Now comes Roche, a major diagnostics company with extremely deep pockets - and apparently the patience to spend over a decade and two major acquisitions in developing a platform.  Roche also has the potential to be fully integrated, with other acquisitions made long ago supplying sample prep and all the informatics you might want. And beyond deep pockets, Roche has global sales, marketing and distribution networks, experience with molecular biology logistics and a strong reputation in the clinical diagnostics space. 

Still, it isn’t Illumina.  It also has a bit lower raw accuracy than the current XLEAP chemistry, though the simple duplex kit would seem to solve that.  The question of instrument cost plays highly here; as I detailed in the unveiling piece I expect SBX to be had for far less than a NovaSeq X and perhaps less than a NextSeq 2000.  I would still expect the MiSeq series to be substantially less expensive to buy and therefore a great starter sequencer for small and new labs - but with few barriers to switching such small labs might “graduate” to SBX not NextSeq.

For big data producers, Illumina has touted their 25B flowcell that can deliver 8 terabases of data in 48 hours.  SBX can deliver a similar amount in twelve hours of running, though it isn’t clear how many batches would be required due to ambiguity over maximum run time.  Still, if that were three batches each at three hours, the SBX could finish sooner - and has much greater flexibility in batching.

I’m long overdue to cover the new Constellation technology, but that’s still early access only and critically limited to one sample per lane.  

So Illumina should be very active in devising a countering strategy, as SBX could start snagging customers for their premium NovaSeq X and probably also NextSeq 2000.  Plus Roche has deep experience in the clinical space, and so could be very quickly reaching for high value applications there.

Illumina also just teased they are launching a spatial platform and already have their own single cell kits. Illumina will always tout their comprehensive downstream analytical solutions - but Roche has been acquiring bioinformatics pipeline companies for years, so this won't be a unique advantage.

Ultima Genomics UG100

Ultima is probably the most at risk of being seriously upended by SBX; for starters Ultima doesn’t have the protection of being a market leader.  Ultima has inked a number of very high profile deals for enabling counting applications - the Olink data generation for UK Biobank, another with Chan/Zuckerberg Institute and so on.  SBX is very well suited for counting applications, with a midi read length that can power across any modestly long inserts to acquire barcodes on both ends and counting applications often have significantly less restrictive base accuracy requirements.  

For such applications, read counts are generally more informative metrics than total bases.  UG100 generates 6-8B reads per wafer every 12 hours, with two wafers going simultaneously.  UG100 also has internal automation - a jukebox like system - to enable wafer addition and reagent loading to occur on a regular schedule, avoiding tightly tying employee schedules to run times as with cartridge-type systems such as everybody else.  

But when SBX run times are measured in small numbers of hours, that continuous running advantage erodes.  With both wafers constantly going and 12 hour run times per wafer, UG100 can crank out about 24-32 billion reads per day (rated at 6-8 billion per wafer); at 5 billion reads per hour SBX can keep up or beat that in a day.  Again, details on how exactly that would work in terms of feeding - would SBX require a sensor swap every 2 hours - remain to be worked out.  But very workable.  

Ultima seems to have gotten some traction with their ultrahigh accuracy ppmSeq chemistry, and that Q60+ accuracy is certainly far above what duplex sequencing on SBX is giving.  How much the market will demand high accuracy is an open question - or put another way, how big a market exists and how will it be carved up between Ultima ppmSeq, Element Q50, PacBio Onso, and generic duplex sequencing approaches?

Element AVITI

The impact of SBX on AVITI will largely come down to those undisclosed costs; the closer an SBX purchase is to getting an AVITI the tougher it will be for Element

Element does have some weapons to fight back.  The Trinity on-instrument hybridization capture library schemes are a great concept, enabling labs to easily add hybrid-capture to their repertoire without extensive training or radically different protocols.  There’s also the Q50 chemistry - if there is a market for high accuracy short read sequencing then SBX may not be participating - though said market must be shared with Ultima ppmSeq and PacBio Onso.

Element’s AVITI24 spatial scheme is a direction that SBX can’t follow, as it takes advantage of AVITI’s optical system to provide high content imaging of tissue culture samples.  If Element is correct that such assays can compete for pharma R&D dollars in the hit-to-lead assay space, then Element will do fine.  But that’s a big if, and dependent on how quickly Element can provide target panels to cover high interest biological pathways.  I’m looking forward to catching up with the AVITI24 team at AGBT and seeing how they’ve progressed since unveiling this approach last year.

Singular Genomics

Singular Genomics was recently taken private by Deerfield Management  Company and so has more capital to stay alive and develop their spatial platform.  Singular has barely a toehold in the market in terms of placed instruments, but a major selling point was the flexibility of the flowcell scheme and the significantly faster run time than competing platforms such as AVITI.  SBX’s fast data generation leaves Singular in the dust, so the question of whether the remaining market for Singular sequencing survives is largely a question of that undisclosed SBX capital cost.  It may also be that enough core labs and sequencing CROs acquire SBX to offer on-demand sequencing so that few labs will see a requirement to have their own small sequencer.

So Singular’s best bet - and what I assume what Deerfield saw as worth acquiring - is their spatial platform.  Spatial remains a turbulent field - killer apps haven’t yet emerged - and really a series of submarkets depending on instrument specs. 

PacBio - Onso

If you buy into PacBio’s thesis that Onso’s high accuracy is a serious value driver, then SBX should have little impact.  However, it doesn’t seem like the has exactly lit a fire under this idea - the only time you hear “lit a fire” in this context is in regard to the amount of capital PacBio has poured into their short read platforms.  The high quality market must also be fought over with Element Q50 chemistry, Ultima’s ppmSeq and various duplexing library kits.

If PacBio brings the high throughput version of Onso to market - enabled by their Apton acquisition - then perhaps they could be more exciting in the high accuracy space.

PacBio - Revio/Vega

PacBio’s HiFi platforms - Revio and Vega - probably won’t see huge impact from SBX, though of course research and corporate budgets are ultimately zero sum games and so every dollar spent on SBX won’t be a candidate for HIFi.   If Roche prices SBX very aggressively, then it would again widen the cost differential between HIFi and short read human genomes and whether the improved information yield from HIFi is worth the premium.  

Many applications for HiFi do rely on read lengths longer than the 1200 basepairs that SBX effectively tops out at and are tuned for the high accuracy delivered.  But PacBio, like ONT (see below), is vulnerable for applications such as whole transcriptomes which stray into the “midi read” zone which SBX can operate in.

Oxford Nanopore

While in the prior piece I used Oxford Nanopore’s technology as a reference point for explaining SBX, the performance of the platforms is so different that ONT might be the least affected by the new platform - other than constantly having to explain how they are so different.  Roche’s instrument certainly won’t fit in your pocket, can’t perform direct RNA sequencing, can’t detect base modifications on native nucleic acids and most certainly won’t be generating 10 kilobase reads, let alone 100 kilobase or megabase sized ones.

One speed, ONT still has a huge edge on time from extracted/amplified DNA to first data.  But if you want a lot of data in a hurry, then Roche has an advantage - though technically if money is no object then with a PromethION P24 running 24 flowcells in parallel you could make it an interesting race, as those 24 flowcells could generate a lot of data before the ~5ish hours to start running on SBX would be over.

The one place where SBX might crimp ONT - besides just sucking up attention and laboratory budgets - is in what I’m calling midi read sequencing.  For things like whole transcriptome sequencing, ribotyping and some other applications, SBX’s ability to get to 1200 bases may be good enough for many users.  As I’ve mentioned multiple times here, there will be a piece outlining the midi read concept tomorrow (I can actually claim that, as it will be in the can before I post this!).

Complete Genomics

Complete has the persistent challenge of being under the crosshairs of US competitive policy such as the Biosecure Act.  Who knows what will happen with that in the current chaos in the US capital, but no clarity or resolution is likely.  Complete hasn’t been completely stymied in placing instruments, but it is certainly a giant headwind.  But even without that, SBX is a clear threat to Complete’s high throughput devices.  

Ion Torrent

A piece I meant to write last year was going to be titled something like “Ion Torrent: I’m Not Dead Yet!”.  Because Thermo Fisher eschews venues such as AGBT and hasn’t breen upgrading the underlying sequencing platform, I’ve tended to ignore it.  But the reality is that Ion has carved a steady business in clinical testing.

Now comes along a serious diagnostics company with an existing portfolio of clinical instruments and tests - and the sales force to go with it.  Thermo shouldn’t be underestimated, but Roche SBX might be the product that gives Ion Torrent a good hard shove towards oblivion.

The Rest of the Ecosystem

We seem to have gotten beyond new sequencing technologies worrying about library structures; Roche says they have a conversion workflow. Though maybe some overdue changes will be induced: in their webinar Roche mentioned that the 10X Genomics Cell Ranger software still wants paired-end reads, so they generate pseudo paired end reads from the SBX reads. Seriously? There's already Ultima plus the long read platforms - time to stop requiring fake paired end reads!

For the other sequencing platform technology startups out there, SBX just makes life harder. Each new platform raises the bar on what a minimal entry can look like in some part of the space, and Roche SBX just claimed a huge amount of "short to midi reads, very fast, decent quality" space.

.

[2024-02-20 6:54pm - completed the sentence Anonymous spotted as a fragment ]