Tuesday, January 09, 2018


Illumina CEO Francis deSouza's J.P. Morgan Presentation did not disappoint.  While humdrum financials and touting market dominance and areas of future growth came first, then came the big Firefly announcement (with a name change to iSeq 100)  -- and then after another short spell of reviewing the latest Nextera chemistry came a smaller bombshell -- Illumina is partnering with former arch-rival Thermo Fisher (nee Life Technologies nee Applied Biosystems) to move the AmpliSeq multiplex PCR technology over to the Illumina platform.

First iSeq 100, as that's what really gets the blood racing -- a new platform.  Also a chance to flay myself for some forgotten cookie crumbs and bad guesses.  And to marvel that such an obvious sequencing technology name hadn't already been used -- or locked down by Apple.

iSeq is the culmination of the Firefly project.  Despite my wrong intuition, it is optical detection using CMOS -- so Firefly made sense as a moniker.  Two-color chemistry is supported on a monochrome imaging platform by changing the dyes on the flowcell.  A and T start out labeled, so the first imaging cycle picks up W.  A chemical step strips the dye from A but not T and adds dye to C.  So the second step detects Y.   Just as in the other two-color platforms, Gs are dark in both cycles.  Most importantly, this means a retention of Illumina's four-terminators-at-once scheme, forcing the polymerase to take its best pick (oops, another bad guess on my part).  So the chemistry should be very, very similar to the existing two-color data in terms of error profile and sensitivity to library complexity.

iSeq uses the patterned flowcell technology to build clusters directly over the imaging elements.  This also means Exclusion Amplification for cluster generation.  ExAmp has seemed likely to be the clustering technology going forward, so hiccups with it (such as barcode switching) must be corrected now.

The box is touted as the smallest Illumina box yet, with a footprint slightly larger than my iPad Pro ---- weighing in around 16kg.  Instrument pricing is $19.5K. Limited numbers will ship at the end of this quarter, with full production rates hit in the second half of the year.  

Once libraries are built, they are added to a cartridge , the flowcell snapped into the cartridge and that is then slid into in the machine, with all additional operation occurring in a single box.  Illumina is boasting that this setup step can be accomplished in 5 minutes.  In my phone conversation with Kevin Meldrum of Illumina, he pointed out that this means the instrument is completely dry and should require no corrective maintenance.  The excitation illumination is generated by LEDs in the instrument, but there are no optics.

The instrument is for lab use; the specs not only specify "Indoor Use Only" but also not to be used above 6500 feet.  So no operating it in a ski lodge or Leadville or in Quito.  Presumably low pressure throws issues into the fluidics.  

Rated output is 1.2Gbases of 2x150 data in 17.5 hours, with 4 million reads per run -- along the lines of Chris Mason's tweeted out results. Other modes are 1x36 or 1x50 in 9hours (for 0.14Gb or 0.200 Gb), 1x75 in 10 hours (0.3Gb)  or 2x75 in 13 hours (0.6Gb)   Illumina emphasizes that these are conservative specifications and that early users (such as Mason) are frequently seeing higher yields.  The 2x150 cartridges are priced at $625 each, with small (~5%) discounts for packs of four. Cartridges are shipped on dry ice and either thawed overnight in a refrigerator or speed-thawed in a water bath. Cartridges and instruments communicate via RFID tags.

In deSouza's talk, he touted several potential avenues for future performance gains.  He noted that in Illumina's labs much faster cycles -- perhaps 50% faster -- are possible.  He also noted that higher-density CMOS sensors are being trialed and suggested throughput could push 10 Gigabases with those chips.  All of the electronics are in the single-use cartridge, so instrument upgrades should just be consumable upgrades.  That's the philosophy which Ion Torrent and then Oxford Nanopore have pushed, with varying success, as each competitor has ended up developing multiple boxes within their lines that use incompatible flowcells.  But in any case, Illumina sees iSeq as a long-term platform with plenty of growth in its future.

deSouza also emphasized that the Nextera DNA Flex technology, which functionalizes beads with Nextera reagents, will be extended to supply even more sample prep in a single kit.  The initial kits enable library preparation even in the presence of potentially troublesome contaminants such as blood.

The final bombshell was the announcement that ThermoFisher is partnering their AmpliSeq multiplex PCR technology with Illumina.  AmpliSeq apparently has some of the best performance of available multiplex technologies (I lack experience with any modern high multiplex scheme), particularly on low quantity sample or degraded samples such as FFPE.  AmpliSeq seemed to be the one remaining serious edge that Ion Torrent had over Illumina, and now ThermoFisher is giving up there.  Undoubtedly some customers are committed to Ion Torrent, but this announcement certainly won't help recruit any new ones. 

The announcement has been covered on some other channels: Matthew Herper has a piece at Forbes and NextGenSeek has a writeup as does Bio-IT World.

I think I'll stick to my earlier prediction that this instrument will be popular with labs wanting relatively quick, inexpensive validation of synthetic biology constructs.  Some of the third party library kits get library prep down in the $20/sample range and 1Gb is the right range for cosmid/small BAC-sized constructs.  If you had 96 to run, the iSeq cartridge adds only another $6.5 to that -- so less than $30/construct.  For large constructs, that's beyond competitive with Sanger sequencing -- and far less work.

But AmpliSeq brings in an interesting option for gene panels.  Meldrum stated that Illumina thinks that qPCR users running small mutational panels can be convinced to move up to small sequencing panels using AmpliSeq.  Illumina believes the accuracy of their platform holds an edge for mutational panels over MinION, though they acknowledged that for some applications long reads give advantage.

Illumina also suggested infectious disease surveillance as another application, with the Broad's Pardis Sabeti quoted in some of their materials -- her group performed important sequencing during the West African Ebola outbreak a few years ago.  That market could end up a real dogfight between Illumina and Oxford Nanopore, with perhaps some Ion Torrent holdouts -- as each platform holds different strengths and weaknesses in terms of portability, read length, multiplexing, data quality and time-to-result.

And who knows?  With the complexity locked in the cartridge, a well-heeled school might even let undergraduates run one. 


Matthew Miller said...

Keith: Thanks for the info. Where did you see cartridge pricing information? I couldn't find that in the press release. As Elaine Mardis said in the Forbes article, it is the reagent cost that matters more than the hardware cost. That 625 price for the PE150 is a nice spot for amplicons and molecular ecology.

Keith Robison said...

Cartridge pricing was from my phone conversation with Illumina's Kevin Meldrum

On Twitter I've seen another commenter argue the same lines as Elaine. I need to tackle in a full post, but whether that line of argument is valid IMHO depends on the project. I utterly crave low $/bp for my big projects -- but have other projects that are time-boxed and $/run ends up being the key metric.

Anonymous said...

Looks like Illumina are looking to be acquired? Not sure SBS chemistry can be optimised any further.