Wednesday, January 14, 2015

Illumina's Expanded Lineup

In my JP Morgan predictions for sequencing platforms, I didn't do badly.  The only major player to make a platform announcement was Illumina, and they did indeed announce instruments that are not radical departures from the prior platforms.  I am kicking myself for not making more specific predictions, as the nature of the new boxes was really unsurprising and it would have been nice to nail that.
The least innovative of the new boxes is the HiSeq X5, which is just half an X10 system.  So that's purely marketing and pricing -- it doesn't sell for half the price of the X10 and the reagents will cost more. Smart move -- while Illumina just sold another X10 system, there could easily be groups that can't quite go for that level of investment (both the upfront capital purchase and then feeding the ravenous maw with sequencing libraries) but will bite on an X5, which can always be upgraded.

The NextSeq 550 adds array scanning capabilities to the NextSeq 500.  That will probably be popular with core labs that have low volumes of business for either.  Conversely, I would expect that if you have high volumes of both you'd probably prefer dedicated instruments.  In any case, it underscores that we (and I have sometimes been quite dismissive of the future of arrays) who have been predicting the demise of the array market continue to be wrong/premature.

The dual flowcell HiSeq 4000, and its single flowcell sibling HiSeq 3000 are logical follow-ons, bringing the patterned flowcell technology from the NextSeq and HiSeq X instruments into the standard HiSeq scheme.  With a rated maximum throughput of 1.5 terabases, this is quite the beast for sequencing.  

One complaint in the comments on my prior piece, and also stated by Brian Krueger in his blog post on the new boxes, is that this annual upgrade cycle is hard on customers who are buying the boxes.  Many feel squeezed between customers demanding the top technology and funders who really don't like buying expensive hardware each year that loses the "new sequencer smell" in a short span.  It's an issue I easily forget, since I rent essentially all of my sequencing machines.  I also wonder how much one would like the alternative scenario: no improvements in sequencing technology or throughput. Perhaps this cycle will encourage more labs to forgo owning an instrument, which in many cases would probably not be a bad thing; a few well-run sequencers may be better for the community than a multitude of underutilized ones.

There's a clear odd pair of boxes in the Illumina lineup now: MiSeq & HiSeq 2500, as neither has the patterned flowcells [edit: Shawn Baker pointed out in the comments that NextSeq lacks patterned flowcells as well].  Presumably the 2500 will be phased out, since most of its functionality can be had on a NextSeq or the HiSeq 4000 -- though 2x250 sequencing seems to have not made it to the new boxes.  Nor does the HiSeq 4000 seem to have 2500's rapid run mode, though I haven't seen a really clear comparison of the time, output and read lengths of the two boxes.  The spec sheets also imply that some library kits (such as Nextera and TruSeq Long Synthetic Reads) may not be supported on these new machines, though perhaps that is just my confusion with the hydra-like growth in Illumina library branding.  So perhaps HiSeq 2500 has a bit of further life.

MiSeq is the anchor at the low end of the Illumina lineup, with the lowest cost of entry and also economical for running.  In addition, it is the only family member to support 2x300 sequencing, which is very handy for amplicons and small genome sequencing (and probably many other things).  MiSeq is also the longest serving model name.  So it would seem likely that a second generation MiSeq will be in Illumina's next batch of instrument announcements and perhaps that announcement might be before a year from now.  I would expect a new MiSeq to have patterned flowcells, the major difference between MiSeq and the newcomers.  

Alternatively, perhaps Illumina will try to slot in an instrument with an even lower entry point.  For example, what if the NextSeq 2-color chemistry was moved into a box even smaller than a MiSeq, with a more limited capacity than MiSeq.  My guess is that MiSeq sells near cost (at $100K, it is roughly the cost of that the open source Polonator went for) and the NextSeq's simpler optics were supposed to enable cost savings, so could the price be shaved down to sub $75K or even approaching $50K?  In addition to getting more small labs onto the Illumina bandwagon, a mini-MiSeq with fast cycling might have unique capabilities for biosurveillance, infectious diseases and other applications in which time is critical.  Many of those applications can get by on much smaller datasets than the MiSeq gets in its fastest mode (a bit over 0.5Gb in 4 hours).  

If you'll indulge some really wild speculation, could Illumina's chemistry be moved into a context to eliminate the expensive optics and make a really inexpensive box?  Since the patterned flowcell technology enables great control on spot geometry and spacing, would it be possible to have the optical sensors mated directly to the flowcell or perhaps even embedded within it?  Even if the number of active spots went drastically down, if this enabled a box to be sold for $20K it could enable entirely new markets for sequencing.   Perhaps something along these lines is what Illumina is working on for their "sequencer on a smartphone" effort.

JP Morgan brought one interesting announcement from a well-funded new player, but that's a post for another day.


Shawn Baker said...

Unless Illumina just made a major change with the v2 chemistry, the NextSeq does NOT use patterned flow cells. Only the HiSeq X Ten (and now the X Five, 3000 and 4000) use these. The NextSeq uses a comically large flow cell that isn't optimized for density (hence the cheap optics) - I don't think patterned flow cells would help here.

Keith Robison said...

Shawn -- thanks for the correction; I shouldn't have relied on memory for that. So I'll predict that NextSeq will get patterned flowcells :-)

Anonymous said...

Hi Keith, could you expand some more about how you rent your instruments?

Keith Robison said...

At both my current employer and my previous one, all of our production sequencing is contracted out to service providers -- in that sense I am a renter not a buyer. Not truly renting, but similar -- neither the costs nor the benefits of ownership

Anonymous said...

So there's a time loss and logistics to manage using a service provider. That's a significant difference from renting and having the instrument on site.

Keith Robison said...

Yes, there is a tradeoff -- you lose control and some speed (for Sanger sequencing in Boston that speed loss in a few hours; for MiSeq it's more along a few days). I should have been more careful in my wording -- we rent sequencing services, not individual instruments.

Brian Krueger said...

Just to follow up on my previous comment (and thanks for the link to my post!), I don't think the 3000/4000 offer anything compelling for a production sequencing lab just yet over the HiSeq 2500. It has greater throughput, sure, but anyone that owns HiSeqs knows that the throughput only matters when the machine is running. Unfortunately Illumina's reliability issues are well known for anyone that has hands on time with the systems. I really see no value add here for being an early adopter.

And to your point about the sequencer release schedule and access to new technology - it would be ideal if 1) Illumina gave us more than a month of warning before a sequencer release and 2) worked hard to be sure there was a reasonable upgrade/trade in path for groups that have purchased their systems recently. It is incredibly hard for me to understand why the 2500 can't be upgraded to run patterned flowcells since that is essentially the only piece of the HiSeqX that is being ported over. I can get 1.2TB of data off of a 2500 run so the extra 0.3GB afforded by the 3000/4000 doesn't represent a "disruptive" increase in data output or justify the release of a $900K sequencing machine on the heels of major upgrades to the 2500.