Sunday, January 10, 2016

Illumina JPM/AGBT Predictions

The JP Morgan conference is next week, which for several years now has been Illumina's venue for making major platform announcements.  So naturally, based on not even anything as substantial as scuttlebutt or rumor, I'll venture some predictions. 

Imagine surveying the sequencing landscape from Illumina's throne.  In the really big project market, such as whole genome resequencing of human individuals or populations, Illumina essentially has a monopoly. BGI launched a challenge -- and then blinked.  Ion long ago abandoned any plans to move into this space, and the technical leadership which might have driven such a charge (no matter how quixotic) has melted away to other organizations.  PacBio making impressive gains in throughput and wowing many with results, is still far away in cost per genome, so as long as conversations stay around that metric (as opposed to more arcane notions of haplotype accuracy or resolving structural variants), it's a clear win for SBS. 

Down at the low range of capacity, particularly for amplicons and small capture sets,  is a bit of a scrum.  Ion competes here, and has a following.  QIAGEN launched their machine into this space, though it would take a serious army of information gatherers to figure out what its capabilities.  The GnuBio instrument is perhaps still being worked on at BioRad, and might be another entry into the amplicon market.  A bunch of folks of demonstrating that Oxford Nanopore can play here, though only small numbers of researchers are doing this and not any big players who could push systems into healthcare systems.

There's two areas which are still not huge, but not covered well by Illumina.  One is long reads; Moleculo (now TruSeq Synthetic Long Read) hasn't fire, with only a smattering of publications demonstrating its use.  The exquisitely clever scientific group published a nifty way to leverage Nextera to yield this information, but there's been no push to commercialize that.  PacBio had a lock on this market, but Oxford (and some noisy beguiled bloggers) make noises.  Startups such as 10X and Dovetail perhaps could be slurped up as a quick entry.

The other uncovered base is very fast turnaround of small assays. Biosurveillance, food safety (Chipotle!), antibiotic testing -- all potentially large markets that are starting to look interesting, but where time-to-result rules.  Often times these assays use small amplicon panels or can glean the key information from relatively modest shotgun datasets.  MiSeq, which is the veteran out in the marketplace for over 4 years, overshoots many of these.  Plus QIAGEN is trying to sell the idea of running many assays in parallel but not in sync, matching erratic streams of samples in these situations.  

A small announcement, and highly likely, is to enlarge the set of instruments with ordered arrays, perhaps simply releasing these for all the remaining instruments.  Since ordered arrays reduce the variability in cluster density, especially in the face of imprecise DNA quantitation, this would enhance the platform as it pushes into markets where the operators may have little molecular biology experience.

How to expand presence at the low end of the market?  How about a MiniSeq?  Use the less expensive NextSeq optics and chemistry to build an instrument with smaller capacity (and ideally footprint) than a MiSeq.  NextSeq uses 8 imaging units; could a machine using only one imager find a niche? Could a $50K price point, perhaps selling around cost, be achieved?

Quick et al have already demonstrated that the existing MiSeq chemistry can support 1x75 reads urs by trimming various times and reducing the number of imaged tiles on the flowcell. NextSeq-style chemistry means only 2 scans per cycle, enabling faster imaging. Depending on the output needs of the assay, even faster times might be possible by further trimming the number of tiles imaged.  Perhaps even faster chemistry cycles could be enabled using a smaller, more streamlined flowcell.  Of course, also potentially on the horizon is new chemistry using much the same hardware, which a proof-of-concept was published earlier this year.

Would a MiniSeq be a single box, or perhaps instead a modular set of linkable boxes capable of running asynchronously, in order to compete with QIAGEN?  Might compute or even reagent coolers be shared?  Or perhaps independent machines, but designed to be stacked to conserve on space?  On the other hand, the heat generation from these machines I'm told can be considerable (due to the lasers), so perhaps completely independent boxes will enable gangs of machines to be placed where a customer's HVAC can best handle them.

A true rabbit-out-of-the-hat would be a radical new chemistry, perhaps a single molecule approach such as Illumina has been long rumored to be working on.  Even just outlining a commercialization strategy, no matter how likely to be delayed (has any sequencing company ever met a long-range forecast for launching a platform?), could be useful in distracting groups contemplating a Sequel.

Is any of this on target or just the delusions of an armchair business strategist? Between JPM and AGBT, at which Illumina CEO Jay Flatley is speaking, I expect to soon be able to score myself against reality.

I drafted this last night, but fatigue & a router-killing power failure halted forward progress. Today, I became aware that a major Illumina announcement is scheduled for 5pm,
 with an investor conference call to follow at 6pm (both EST). I'm still going to push this out, but a conference call implies something much larger than what I've dreamed above, such as a major acquisition or alliance.  


Anonymous said...

Just received an email from Illumina (10:10 AM EST - 01/11/2016) announcing a webinar on Tuesday, January 19th.

"Introducing the MiniSeq™ System: Our simplest, most affordable sequencing solution yet."

Anonymous said...

Unknown said...

Amazing prediction. Great job.

Anonymous said...

"The exquisitely clever scientific group published a nifty way to leverage Nextera to yield this information, but there's been no push to commercialize that"

Hi Keith, thanks for running this informative blog. I wasn't aware of the use of Nextera for synthetic long reads. Could you please provide more detail about which group demostrated this?

Unknown said...

Wow, you nailed the name! Good job--and definitely ID'd the directions Illumina seems to be looking toward.

Keith Robison said...

Anonymous: it was Illumina in collaboration with Jay Shendure's group that developed CPT-Seq (Contiguity-Preserving Transposition). Two papers launched it: Adey et al and Amini et al. The key obsevation is that Nextera transposase hangs onto the DNA, so if you barcode the adaptors on the transposase, tagment and then dilute prior to a round of barcoding PCR, then reads with the same PCR barcodes + same tagment barcodes are more likely to be from the same input fragment. It is statistical (i.e. many pairs won't be from same input molecule), but if you get enough linkages you can infer physical linkage -- much like traditional genetic mapping.