Roche Nanopore Expectation Management
A Roche Nanopore sequencer appears to be coming.
Yesterday we summarised what we know, and speculated a little.
Today we’re going to workshop some potential scenarios1.
It should be noted that we already know it’s a 4h run generating 15B reads, this is taken as a given.
Any Data
Any data AT ALL will be a huge deal. We waited decades for nanopore sequencers to produce sequence data. We still only have data from one approach (stranded sequencing) from one company.
Another approach would help establish nanopores as a fundamental, general purpose detection technology.
It would also show that Roche has done the work to get an end-to-end system up and running. And regardless of the performance this is a positive sign they may be able to iterate on this.
But no immediate threat to other sequencing companies.
Q10 500bp Reads 4h Sample Prep
90% Accuracy2. Better than Oxford Nanopore’s first public datasets.
Are there interesting applications at this performance level? Yes, but I think you’d primarily need to compete on cost. At $1000 it could be interesting for a few applications.
More important is that it’s a solid position to iterate from.
No immediate threat to other sequencing companies.
Q20 500bp Reads 4h Sample Prep
~This is what the rumor mill is expecting~
At least in terms of accuracy, this seems to be the consensus. 4h sample prep is me speculating that sample prep won’t be longer than run time.
The end of Illumina style sequencing-by-synthesis?
Ye…yes?
I can’t see a way that Roche chip cost-of-goods is more than $1000, probably closer to $300, and less in volume3.
The instrument shouldn’t really cost more than a Revio or Vega to make, so at most $300K to $600K.
Chips and instruments should scale down, beating Illumina style SBS across the whole range from iSeq4 to NovaSeq.
Q20 is “good enough” for the vast majority of applications5 and will get better.
We don’t know how Roche will market this. They might decide to just market the platform for clinical applications. They might decide they need to sell kits for $10000.
But it would still be over.
With the method established, one day the IP protection will run out6 or someone will find a work around. Sequencing-by-synthesis’ days are numbered.
Immediate threat to Illumina.
Immediate threat to Ultima/Element.
Immediate threat to PacBio short read systems nobody cares about anyway7.
Significant threat to Oxford Nanopore (they retain direct RNA and long reads).
No immediate threat to PacBio long read systems?
Q30 20Kb+ Reads 1h Sample Prep
As long as they sell it at a reasonable cost basically everyone is screwed.
Roche buys Oxford Nanopore at fire sale price so they can do direct RNA and very long reads.
Not investment advice, and not 100% serious. But it’s fun to speculate!
All accuracy and Q values are assumed to be single base raw read accuracy uncorrected. Not using an overfitted AI model trained on the reference of the sequence being evaluated. An honest to goodness trad single base accuracy. The accuracy you’d get if you take all the reads, align them against a known good reference (created using another platform) then for each position calculate the number of bases that match and the number that don’t match. And then express that as a fraction.
I’m sure I don’t need to qualify this… but you never know!
PacBio currently sell their chips for ~$1000, like Roche they use a standard CMOS process. Roche may need to do some post-processing work to add electrodes (previous Genia work showed Ag/Cl electrodes on device). The Roche chip can likely be smaller than PacBio. Roche only have 8M wells (as opposed to 25M). Roche wells are ~4 micron, making the chip much smaller.
Chips and instruments should scale down a 2.5mm^2 chip generating 1B reads? Much better than the iSeq, MiSeq, i100, NextSeq and similar to the NextSeq 2K. That’s a tiny piece of silicon, far smaller than that in the iSeq cartridges.
Assuming no strong systematic bias.
Perhaps by 2036 going by some of the Stratos patents.
Sorry, I know a lot of people work very hard on these platforms. Maybe they will get more traction one day. But for the moment, they don’t seem to be having a huge impact on the market.