Monday, January 12, 2015

2015: Another Year of Sequencing Evolution (not Revolution)?

The J.P. Morgan Conference is firing up, and for the past few years that has meant big sequencing platform announcements -- HiSeq or Ion Proton or such.  This has stolen some of the thunder from AGBT in terms of major announcements (sadly, I won't be attending this year -- and will try not to land my self into surgery the way I did the last time I didn't attend AGBT).  I figured I'd better write this tonight before any more JPM-related sequencing instrument announcements show up, or more to my prediction, before the conference ends without any.
That's right, I'll stick my neck out and bet that there won't be any major announcements this year at JPM in terms of sequencing instruments.  Illumina has little incentive, as they've pretty well run the table for large sequencing projects with various versions of HiSeq, as well as their MiSeq being the dominant machine for many smaller projects (and NextSeq appearing to catch on as well as the middle sibling in the family).  Illumina just announced another HiSeq X10 purchase, which I suspect will continue throughout the year; the number of really wealthy institutions or companies wishing to sequence entire human populations will certainly grow slowly but steadily. 

Ion Torrent abandoned the field to Illumina for whole genomes last year, throwing in the towel on the PIII chip for the Proton while struggling to launch the PII chip.  Ion's machines have carved out a niche for fast sequencing of amplicons and a few other adherents, but are clearly in also ran status.  I think SOLiD still exists, but one hardly hears about it anymore.

Roche abandoned 454 last year, after being nearly completely overwhelmed by MiSeq (with contributions from Ion and PacBio).  They bought nanopore sequencing company Genia, which has looked promising, but little noise on that front in terms of starting testing in the community -- either it is going to be very quiet or just isn't going to be any time soon.

QIAGEN was supposed to start providing some real competition, at least for the MiSeq and Ion Torrent PGM. Or the Proton? Who knows -- QIAGEN has yet (unless I missed it) released any sort of hard performance numbers, such as throughput or read lengths or error rates, making any estimate a blind guess.  The launch date for their instrument always seems to be well in the future as well.  Alas, I've failed to record the various announcements, as both QIAGEN and Ion Torrent PII could make for very entertaining slipcharts.

Pacific Biosciences has carved out several nice niches, but as long as their machine remains a huge investment they will have difficulty greatly expanding their market.  Don't get me wrong: the amazing technology is oh-so-close to being science fiction and they've seeded the community with all sorts of really cool bioinformatics.  But when the big pushes are for population level sequencing, the throughput just isn't there.  PacBio partnered with Roche for diagnostics, leading to persistent rumors of a desktop PacBio instrument -- which would be breathtaking.  I'd love to fail in my "no JPM" prediction with that baby surfacing, but doubt it.  

Oxford Nanopore hit the world last year with an early access program.  This was recently expanded, with a new round of excited posts of photos of their USB-stick sequencers by new testers.  Nanopore data has even been published in multiple journals, though the company is still grappling with uneven performance of the devices (disclosure: via my employer I am part of the early access program, and also served as a referee on a paper which was published).  Data quality remains an issue, but with steady improvement from well below PacBio to sometimes significantly exceeding it.

Oxford could take on PacBio, but instead has been painting a picture of much larger markets capitalizing on their unusual advantage: a complete sequencing laboratory that could reasonably fit in a small backpack.  With the ongoing Ebola outbreak, the idea of molecular testing anywhere is appealing.  Oxford has also suggested in conference presentations and Tweets the idea of performing most if not all of the library preparation on the flowcell, clearing out many of the items in that hypothetical backpack.  Oxford  has also floated the interesting idea of "run until", in which real-time monitoring of the sequencer enables halting sequencing after certain numbers of events have occurred.  Since their flowcells can be reused (though not indefinitely), that is an intriguing concept.  Oxford has also announced the Borg version of the MinION: a notebook-sized array of 96 flowcells called the PromethION, with each flowcell either acting independently or all on the same sample.  In any case, Oxford is very explicitly ignoring all the old patterns for product announcements, not even tying them to major meetings of any sort.

What else?  There's a host of nanopore companies, but few have (as far as I have seen) even demonstrated running anything serious.  GenapSys made noises about launching an Oxford-style early access program, but beyond signing up recruits has made no significant announcements. BioRad swallowed up GnuBio, which had an intriguing microfluidic sequencing-by-hybridization technology for the amplicon sequencing market, but has gone relatively quiet.

So what do I see through my coke bottle glasses?  Illumina may well continue to evolve their platform, with further tweaks to increase yield.  But as long as they lack a serious competitor, the risk of something going wrong may continue to feed a cautious approach on the market leader's part. In any case, Illumina has shown a healthy appetite for moving into the clinical space, which is almost certainly where the big money lies.  Research markets remain the source of those juicy clinical applications, but are valuable only as springboards. So if there are big announcements from Illumina, I expect it will be another big foray into big applications.  

So, that's my prediction -- what I truly believe, though I might not love it. Time will tell if I'm off the mark -- and my hoping to be so is unlikely to influence the outcome.


Rick said...

You didn't mention the announcement of Illumina's 3000/4000 machines. While far from revolutionary if their per-base cost is low enough they could still drive the purchase of new equipment (and thus Illumina's bottom line).

But, yes, it does seem to be a year of evolution instead of revolution.

Keith Robison said...

I actually wrote this piece about 6 hours before the Illumina announcement, which I would characterize as further evolution of their platform. Revolution would have been completely new sequencing chemistry, or radically shorter run times, or a sub-MiSeq box with a very sub-MiSeq price, or crazy long reads or ...

Anonymous said...

Evolution might be still be a little generous, iteration seems more appropriate.

The issue for customers is that now there are a lot of choices and all with significant outlay, and there seems to be a good chance in twelve months there might be more choice given Illumina's track record.

So, is it wise to wait 12 months or longer? The sales model works well for consumer electronics at a few hundred bucks but this is serious money and lab real estate being occupied.

Brian Krueger said...

The cost per GB on a HiSeq 3000/4000 is on par with that of an over clustered HiSeq 2500. The 3000/4000 isn't worth the investment just yet. Maybe when they introduce the next iteration of their more compact patterned flowcell it will be worth it because that might actually drive down the costs. That's the problem with the patterned flowcell, you're limited in your ability to cluster it so pushing the cluster densities to get more bang for the buck isn't possible on this system. Surely in the future the patterned flowcell will be a great improvement over random clustering, but for now I think it's smart to let the early adopters work out the typical Illumina launch issues. I hear there are still problems with massive cluster duplication on the X, so yet another reason to let this simmer.