Sunday, January 12, 2020

Is Illumina Vulnerable At The Bottom?

According to both the FTC and CMA, Illumina holds a nearly complete monopoly on the sequencing market, with other players (Ion Torrent, Oxford Nanopore) holding on to toehold niches.  Illumina has held that position for an extended period, so what might upset it?  I'm going to explore the case that they may have some serious inroads at the bottom of their line.
The only current contender for competing on short reads at scale would be BGI/MGI, but as I noted last time they seem to have a Xeno-esque approach to entering the U.S. market, -- they're always announcing getting closer but never actually reaching the destination..  Perhaps that is due to serious IP challenges, but one would think between their CoolSeq technology and the QIAGEN GeneReader IP presumably being available for a price that they could solve this.  Not only does BGI threaten the top, but they also announced this fall plans for an inexpensive benchtop sequencer.  However, that won't be beta tested in China until this spring, so it can't be a player anytime soon.  BGI/MGI appears to not be presenting at JP Morgan this year.

Genapsys is actually on the market now with an inexpensive benchtop instrument.  Currently it roughly competes with iSeq in terms of total yield and even uses Illumina libraries.  However, it is a single read per fragment, which isn't as useful for some approaches and certainly nixes any clever approaches that rely on reading both barcodes.  I suspect it can only read in-line barcodes and not the Illumina Nextera-style ones which require additional reads.  There's also the difference between being to read a fixed 150 bases from each end of an amplicon, for example, versus reading 150 bases from one end of a larger number of amplicons.  Plus the difference in error profile, which Genapsys really needs to release more data around.

Genapsys will be presenting this week at JP Morgan (2pm PST on Tuesday).  One of the big points they should be covering is when they plan to launch their 144M sensor chip.  With that, their iSeq-sized box will be producing data well in excess of a MiSeq and actually near the low range of the NextSeq.  Also to watch for, though perhaps not in that forum, would be more data on basecalling accuracy (particularly for homopolymers) and any plans for longer read lengths or paired end reads.

In his SanDiegOmics Blog Shawn Baker lays out a nice description of the current state of sequencing and makes some predictions.  One prediction I wish I had thought of is that the MiSeq has been around a long time, uses much older 4-color chemistry (but does so well with it!) and so Illumina might announce either a replacement or something to go between MiSeq and NextSeq.  It would be interesting to see how that might fit in, but NextSeq is more than 2.5X the price of MiSeq so there is a space there for a machine in the $150K-$200K range.  Or perhaps a MiSeq replacement using 2-color chemistry to speed the runtime.  Or some hybrid thereof?  Perhaps a slightly higher than MiSeq price for an instrument with multiple consumable sizes, one MiSeq-like and one halfway between MiSeq and NextSeq Mid-output (or about 15-20 gigabases in 2x150 mode). Is there a demand for a such a beast, or would it just be part of Illumina's strategy to have a ladder of instruments so that small labs can be consistently upsold into higher consumable usage patterns?  Or just to have a box that covers the same range as the 144M chip from Genapsys?   Illumina presents 8:30am PST on Monday, so any announcement could leak out at any moment.

The other wildcard in this is Oxford Nanopore's renewed push into short reads.  highlighted at the New York Nanopore Community Meeting.  Even with the Mk1C unit sporting a built-in computer, ONT has the lowest capital requirement offerings.  But getting labs to switch to it is the hard problem.  What Nanopore could really use here, in my opinion, are some methods and papers showing how existing short read workflows can be easily modified and used to generate comparable results on nanopore.  Now, I think it important to emphasize this means being fully transparent about errors and such -- no broad claims of Q-amazing accuracy without detailed caveats on error modes (the subject of another embryonic post).  Better yet, pick applications where nanopore's inherent lower accuracy just isn't critical, such as many counting assays.  Or even suck-it-up and show how nanopore might rapidly QC samples prior to loading them on the really big Illumina iron.  Things like checking barcoding balance across a pool of libraries.  But that would require some way to rapidly convert Illumina libraries into nanopore libraries so that the time and effort of conversion is more than paid for by getting fast and informative results off a MinION.

Of course, if anyone is planning to push their startup into the commercial wilds, they'll probably be targeting the low end as well.  There's a lot of companies out there with various levels of visibility into their technology and proximity to launching into the marketplace.  But that's a topic for a post devoted to the subject, which I hope to produce soon.

[2020-01-15 -- fixed MiSeq-->NextSeq mutation per Wouter's comment]

No comments: