Who's Going to Buy Roche Axelios?

After Roche's AGBT presentation, a key question for many is "who is going to buy an Axelios?".  Below are some thoughts on the topic, based around general classes of sequencing labs that exist in the world.

Platform Collectors

There are labs which seem to have a philosophy of "own one of everything". If you ask around, there will be a Broad consensus on the exemplar institution for this pattern. Such labs supply us with cross-platform comparisons, which we will all be grateful to see. Something to watch is which labs try the Axelios and then make it a secondary instrument, and which actually put it in their main lineup of services.  One can also expect a great deal of exploration of the midi-read regime for single cell sequencing.

Academic Core Labs

From a completely unscientific survey consisting of chatting with a few core directors and observing social media posts, it would seem that very few core labs are in a hurry to try Axelios.  These are challenging times for academic cores, particularly in the US as the Federal funding situation continues to be as predictable as a Magic Eight Ball toy.  Even without that, cores are facing aggressive competition from both traditional sequencing CROs as well as the new entrants offering superfast nanopore sequencing.  None of this makes them eager to sink a huge sum into yet another platform instrument. 

Cores, particularly smaller ones, also seem hesitant to commit to Roche until the full implications of Roche's still somewhat ambiguous pricing model are worked in.  Specifically: how much do the advertised prices require a certain level of volume?  For example, the expandomer creation instrument can process one to four flowcells in a run - does the best pricing effectively require running at that level or can the same cost per sample be achieved by running only one flowcell in a day?  And some cores may find that in general Axelios, as well as NovaSeq X and Element's VITARI, overshoot their throughput needs and therefore don't represent prudent investments.

ABRF, the big core lab confab, just kicked off.  I'm not there due to worrying about meeting burnout - AGBT Agriculture will be my third of the year and right now I'm planning three more before the end of August, but part of me wishes I were to hear the scuttlebutt from core directors talking about Roche.

CROs

Speaking of CROs, who will be the first to announce Axelios services?  Some might do it for the publicity, some because in their view Axelios will offer appropriate batching.  But will customers be resistant?  Illumina data is a very well understood entity, and many buyers of CRO services are not NGS nerds.  So there may be some hesitancy out there - in interviews with CROs even Element sometimes gets hit with the "it's not Illumina and that's what our customers want" tag.

Clinical

At AGBT a key Roche clinical collaborator, Edwin Cuppen in Amsterdam, showed that for a wide variety of oncology scoring systems used in his lab - detecting phenomena such as mismatch repair deficiency - Roche and Illumina were indistinguishable.  The clinic is the big prize, and Roche has historically had a strong presence in the space.

Axelios also has an interesting aspect relevant to cell-free DNA.  With existing short read platforms, the size of the flowcell sets an upper limit on the number of reads.  But on nanopore platforms, the library length distribution sets a limit - if you have shorter libraries, then more reads come through. Since cell-free liquid biopsy samples are dominated by very short fragments, the yield (and therefore cost) per sample will be better than what is advertised for WGS libraries.  In a recent LinkedIn post, Alex Dickinson touted this and also Roche using a single read of sufficient length to measure the cell-free insert length - as key advantages over Illumina in this space.  The actual length is important since this contains useful information as to the source and process which generated the cell-free DNA - at least it will if the library prep hasn't distorted it (see my recent Canal Biosciences piece)

In the rapid space, Broad Clinical Labs has already announced they will offer as a service their rapid newborn sequencing method.  The medical and humane value of such a service is clear; how much medical systems will engage it and how much payors will finance it remains to be seen.  If Broad Clinical Labs and Roche succeed here, it will bring up the question of why nobody has succeeded before?  After all, Oxford Nanopore can deliver rapid results as well.  

Machine Learning Crowd

There are some nonprofits or companies that specialize in generating huge datasets for training machine learning models - disclosure/pitch, I work for one. Much of this is done on NovaSeq X, but there's also significant use of Ultima platform in this space, since often these are reading single cell or spatial data and the cost per datapoint is paramount. 

So will Roche have advantages here?  Probably not in general, but again the midi-read space could be very interesting.  If you really want to model splicing, for example, don't you want really deep, high accuracy datasets of what splice forms show up in which tissues?  Being able to generate billions of reads in the 500-1000 basepair range could be very helpful for this.

Coda

Presumably this summer, or perhaps even before, there will be splashy press releases around new Axelios customers.  We'll get a better picture on what categories of laboratories are buying - and if any of what I wrote about holds any water.