Sunday, January 07, 2018

Illumina Outlook II: The Fleet

In my prior installment I looked at Firefly, now clearly a working instrument.  Now I'll take a peak at the rest of the Illumina fleet.
Illumina now offers a wide array of sequencing instruments, ranging from the  MiniSeq to the monster throughput NovaSeq.  For the latter, Illumina launched the S4 chip last year which offers dizzying amounts  -- up to 3 terabases -- of data.  Firefly will push further into the low end of the market. 

In the research market, Illumina has essentially swept aside the competition.  Ion Torrent still has a toehold, but not much more.  SOLiD is gone as is 454.  QIAGEN has their GeneReader, but focuses exclusively on the diagnostics market.  SeqLL keeps the Helicos technology running; more on this soon.  BGI-Seq boxes are in the wild -- but not (as far as I can tell) in the U.S. and haven't made an impact. Pacific Biosciences and Oxford Nanopore both offer very different sequencing devices with very different capabilities, but neither is challenging Illumina in most markets -- but stay tuned there.

While the different boxes have variations from the original theme -- some use patterned flowcells with exclusion amplification for cluster generation and all the newer boxes have used two-color imaging -- in terms of final data they are mostly the same.  Two color labeling can lead to problems on low complexity samples, such as amplicon datasets, and exclusion amplification has led to some well-publicized high rates of barcode switching.  I've also heard whispers that exclusion amplification can lead to higher rates of cluster duplication.

Illumina doesn't always have the Midas touch.  The NeoPrep library preparation instrument was abandoned last year due to incurably erratic performance.  Linked reads are hot, but Illumina bought an early technology called Moleculo that works on short ranges (5-10Kb) and not the very long ranges (>50Kb) where the greatest interest lies.  Illumina's research groups have published several schemes for linked read generation, but none of these have materialized as launched kits.  Multiplexing with Illumina kits seems to be stuck in the midrange of about 96 samples.

There's also the variation in read lengths.  Paired end 2x150 has become the default on Illumina platforms, which is impressive for a scheme that launched at about a sixth of that.  MiSeq supports 2x250 and 2x300 formats, though the 2x300 has attained a reputation for erratic performance that sometimes veers into unusable (alas, I can personally attest to that -- gave up on the 2x300 chemistry after a serious disaster).  HiSeq 2500 Rapid Run appears to be only other platform supporting 2x250.  

So an easy brand extension for some of the boxes, perhaps NextSeq and/or MiniSeq, would be to enable 2x250 chemistry.  That would boost throughput, but more importantly enable those amplicon applications that need longer reads. Of course, to date only four-color instruments have gone beyond 2x150, so that may be challenging.  Or Illumina could commit to fixing the 2x300 MiSeq chemistry

MiSeq was announced seven years ago and is the oldest instrument in the fleet.  With a diagnostics version on the market, it is likely to stick around for many more years.  MiSeq also has a larger number of run formats than its little cousin MiniSeq, with overlapping performance.  For example, the MiniSeq Hi Output 2x150 is actually rated for nearly 50% more data than the MiSeq v2 2x150 chemistry.  This is because MiniSeq yields as many clusters as the v3 chemistry on the MiSeq.

So one easy way to broaden the utility of each instrument would be to launch additional kits that extend the ranges to overlap more.  MiniSeq and MiSeq (and soon Firefly) overlap in output but with differences in run time and format.  If someone already owns a MiniSeq, convincing them to buy 2x250 kits is probably easier than convincing them to upgrade to (or expand with) a MiSeq.  A 2x250 mode on MiniSeq would be projected to yield about 10.5 gigabases of data, which is actually better than the MiSeq v2 2x250 mode (again, because of higher cluster density).

Another version of line extension via reagent kits would be to over lower capacity kits -- similar to the rapid kit on the HiSeq 2500 or the Nano and Micro kits for MiSeq -- offering lower cost-per-run (but higher cost-per-base) and faster runs (due to less scanning) than the full size kits.  MiSeq v2 Nano delivers only 4% of the data of a full v2 2x150 run, but does so 5 hours faster.  If that supports a quicker design-build-test cycle or your assay just needs a bit of data, it could be a win.

Propagating exclusion amplification clustering and patterned flowcells, currently found only on the HiSeq 3000/4000, HiSeq X and NovaSeq, is potentially another line extension.  Exclusion amplification enables higher cluster densities, albeit with the barcode switching (and potentially duplicate cluster) issue mentioned above.

Or perhaps Illumina's managerial energies are mostly focused on their more downstream adventures such as Helix and Grail?

We'll find out more on Monday, January 8th when Illumina's CEO Francis deSouza presents and gives Q&A at the J.P. Morgan conference, at 3:30 and 4:00 PST respectively.  Word on the street is that Firefly will get strong coverage in the presentations -- perhaps monopolizing the whole affair.  To be followed soon afterwards with further kibitzing in this space...




1 comment:

James@cancer said...

I think there's a real market for low cost runs...the difficulty is that filling the cartridges, making the flowcells, putting the whole lot in a box and sending it around the world costs the same whatever the yield. But with bog margins on reagents I think Illumina could swallow a low cost kit that enable faster and cheaper method/experiment development. However if users are going to run on the current systems, or buy a new box that is "cheaper to run" then why release low cost kits?