Roche Axelios1 Run Pricing: Not A Knockout - Yet?

By far the most awaited news at AGBT was Roche's reveal of the pricing scheme for Axelios 1 consumables: what would be the structure and how much would different runs cost?  Indeed, Roche sequencing systems chief Mitu Chaudhary walked on to "The Final Countdown".  Roche did show its hand here, but stayed coy on the launch schedule beyond "Summer 2026".  Roche's pricing scheme clarified many questions, but also raised new ones.

Beyond the actual number, the key part of the announcement was the pricing structure: Roche will have a single kit good for 4 hours of sequencing.  So no exotic pricing such as metering the number of reads or gigabases, nor any special pricing for kits for the rapid results market.  A run is priced at $2400, which for duplex genomes works out to $150 a 30X genome.

That number has been viewed by many, including some nervous competitors, as not a knockout blow.  Element is promising $100 a genome with VITARI, Swiss Rockets/Complete Genomics also $100 for CoolMPS 600, Illumina's new 35B flowcell (similar to what many labs are getting from 25B, so perhaps old wine in new bottle) perhaps also around $150, and Ultima had promised $80 with Solaris 1.0 so likely lower for the 20B wafer on Solaris 2.0.  

There's still a bunch of hidden math in Roche's calculation.  Another key reveal is the pricing is based on 20 reuses of the sensor chip, reuse that is completely automated on the sequencing instrument.  This also clicks with the ability to queue 4 libraries on the sequencing instrument for convenient shift operation.  Each run takes 4 hours plus 1 hour to refresh the sensor, so 4 standard runs fits neatly into a simple schedule - though there is no provision to queue up runs for over a weekend or holiday the way Ultima does.  And that 20 number works nicely as 5 days worth of 4 runs each, so a lab running flat out Monday through Friday goes through one sensor per week per instrument. 

Interior of synthesis instrument: where the expandomers expand

But how much is that sensor?  Roche didn't say, which means we can't estimate how much could be shaved off if the sensor reuse was extended further.  Roche did present data showing that a sensor failing before 20 runs - 1 failed at 19 - was rare.  But they didn't say what would happen if a sensor failed sooner - is the cost of that on the user or on Roche?  

Roche did emphasize the simplicity as a feature - clearly a knock on Illumina with a wide variety of flowcell sizes and running kits.  

There were also hints that the simple scheme may have some wrinkles - the synthesis reaction may need to be dialed in for the length distribution of the library pool.  This also means that if two, three, or four library pools are being run in parallel on the synthesis instrument, they must have the same size distribution - can't run a short counting library and a 1 kilobase library in the same run.

Roche also mentioned that users purchasing multiple sequencing instruments would likely not buy a one-to-one ratio of synthesis instruments.  Since both synthesis and a standard single sequencing run are 4 hours, a synthesis instrument could easily supply two or three sequencers.  However, it did not appear that one could feed new libraries into the sequencer once runs had started 

Back to the costing.  Roche said a full 4 hour duplex run delivers 1.8-2.7 Tb of concordant duplex data, but that doesn't include the simplex tails that can assist in mapping the duplex data.  So Roche is rating this at 16 genomes per duplex run, but I wouldn't be surprised if labs start trying to push that a bit further since the simplex data helps the duplex data.  

But Nava Whiteford has already pointed out that if a duplex genome is $150, then by his calculations a simplex genome is $30 - far below any competitor.  Certainly an interesting independent evaluation could be to compare the overall variant calling error rates for a simplex genome and a duplex genome.  It may well be that for many non-clinical applications - particularly for generating overall diversity information from populations or for low-pass genotyping, Roche will support very, very low costs.

It's also important to remember that Roche's nanopore technology, like Oxford Nanopore's, does not deliver a fixed range of reads per run.  Given an ideal library, the number of reads generated is a function of the length of the inserts.  For simplex reads, Roche gave guidance of 40B reads if a library is centered on 175 bp, but only 10B reads for a 500bp mean insert length.  Previous presentations had suggested more than 1B for libraries in the kilobase range, but perhaps that is an underestimate.  What isn't covered here, but could be very interesting to see how productivity might go up with extremely short reads.  If you are sequencing amplified short barcodes or need only a limited amount of 3' RNA-Seq data for fingerprinting an expression response, perhaps your libraries might have inserts of well under 100 basepairs.  It would be very interesting to see the productivity numbers for such short libraries, which might really pinch competitors.  Already, Roche is pointing out that 40B reads for $2400 is 6 cents per million reads - and they like to put extra inflection on "list" when describing the pricing.  Roche also mentioned early results on single cell RNA data, with 20B reads coming in just the first hour.  Roche is partnering with Watchmaker Genomics on their TAPS+ chemistry to provide methylation sequencing on Axelios 1.

On the clinical front, Roche's collaborator Edwin Cuppen in Amsterdam showed that for a wide variety of measures his lab uses for clinical cancer samples the Roche results and Illumina results are fully concordant. Now these are not individual variant calls but rather overall measures to assess properties such as disrupted error repair pathways, but it does testify to the readiness of the Roche platform in key clinical areas - and Roche already owns Foundation Medicine in this space.

Roche pointed to some of the typical places for future productivity to come - faster translocation of libraries through the pores, even denser sensor arrays.    SBX guru Mark Kokoris walked on to "The Best Is Yet To Come", so they definitely think there is plenty of room for further improvement.

One spot where Roche did seem to have a weakness is in the length and complexity of the duplex sequencing library prep - about 9 hours.  They promise this can be automated on standard liquid handling instruments, enabling overnight prep.

So Roche has announced pricing that on the surface seems very conservative and not something to terrify their competitors - but also offers hints of opportunities to be aggressive in the future. For example, it could be that after a successful staged roll-out to already identified customers then Roche would engage in much sharper-edged competition on price.   Competitors also now have an opportunity to respond to the current pricing model - no market is static.  ASHG is about six months from now and will be a very interesting time to see to what degree Roche's commercial strategy finds traction and who the early customers will be beyond the sites that have generated public data so far.