Saturday, February 22, 2020

MGI Dual Drop of CoolMPS News Ahead of AGBT

Friday morning I got excited because a preprint showed up at BioRxiv detailing the CoolMPS sequencing technology from MGI (aka BGI aka Complete Genomics).  First announced in Fall 2018, this approach sounded, well, cool.  Using fluorescently labeled antibodies specific to each reversible terminator seemed like a crazy pipe dream.  So getting a good look at it in a manuscript is an event!  But then Friday afternoon MGI had a second big pre-AGBT reveal: launch of their sequencing systems in the U.S. later this year. Below is a quick run-down of the sequencer announcement; the pre-print has many details I'm still parsing.

I've been guilty of not covering MGI very well.  I don't like ending up parochial, but the lack of their systems in the U.S. market meant I really had little opportunity to see or use them.  So the launch announcement is a needed kick in the seat of the pants. 

The launch is scheduled to begin with the benchtop G series in Q2 followed by the NovaSeq competing floor unit T7 in Q3.  The tiny "sequencing cube" E series is not included, which is unsurprising since it hasn't launched yet.  

The DNBSEQ-G50 is rated for 300M reads delivering 15Gbp 1x50 in 15 hours, 13Gbp  1x100 in 24 hours or 2x50 in 26 hours and 60Gbp 2x100 in 48 hours.  I don't have pricing information to perform a proper comparison, but contrast that with the new NextSeq 1000/2000 P2 reagents rated for 400 M reads and 40Gbp of 2x50 in 13 hours,  80Gbp 2x100 in 21 hours or 120Gbp in 29 hours.  Even if the consumables price per basepair comes out the same, Illumina is delivering more data in a unit time allowing more runs per instrument.  The older NextSeq 550s  in high-output mode are also rated at 400M reads and can deliver 30Gbp 1x75 in 11 hours, 60Gbp 2x75 in 18 hrs or 120Gb 2x150 in 29 hours.  So MGI will presumably be aggressively pricing the sequencers and/or the kits to grab market share, and they claim the CoolMPS technology has much lower reagent synthesis costs so they may have extra cushion to do so.

Step up to the DNBSEQ-G400 sequencer and you get a dizzying array of standard runlengths - 1x50, 1x100, 2x100, 2x150, 2x200 and a long 1x400 format.  These can be run in a standard or "fast" format.  It's a benchtop machine -- but one with a larger physical format than many of the others on the market.  The two big read formats take over 4 days -- 108 hours for 2x200  and one more for 1x400, with both rated at up to 720Gbp, with up to 1800 reads per flowcell.  That's in between the rated Gbp generation for NovaSeq S1 and S2 flowcells in 2x150 mode, though even the bigger S2 will finish in 36 hours and deliver far more reads.  

DNBSEQ-T7 is the NovaSeq competitor, though running 4 flowcells simultaneously versus the 2 slots on the NovaSeq.  MGI claims up to 6Tbp of 2x150 data in 24 hours; there's also a 2x100 mode to generate only 4Tbp in 20 hours.  Contrast that with 3.2Tbp of 2x150 in 25 hours on a NovaSeq S1 flowcell.   It's a huge floor model, looking like it takes up about 50% more floor space than a NovaSeq and definitely towers higher. 

When MGI would launch in the U.S. has been a looming question for some time.  Success could mean real competition for Illumina, driving prices down and innovation upwards.  But MGI hadn't tipped their hand, despite raising $200M last May explicitly to support launch in the U.S.

There are two already apparent wildcards.  One is the tragic coronavirus epidemic raging in China.  Obviously uncertainty in a commercial launch is trivial compared to the terrible human suffering right now, but it is bold of MGI to plan this when their employees and likely employees at many key parts of their supply chain are in the areas hit hard by COVID-19 and the Chinese quarantine measures intended to contain it.  The other wildcard is Illumina, who only yesterday filed an injunction request  to block the import of MGI sequencers and chemistry into the U.S..  The patent landscape around this is a rabbit hole I will absolutely evade at this time. In their response, MGI claims that Illumina's action does not concern the CoolMPS technology but rather is an attempt to block comparing MGI's current chemistry with the CoolMPS chemistry.

So how well will MGI do in the U.S. market?  

For existing Illumina customers there are substantial switching costs. Even if the new sequencers can take advantage of all the existing Illumina library prep schemes (on the list of questions to grill MGI with at AGBT!), there's still a new template prep scheme which will require new training -- and for the high throughput labs new automation and LIMS work.  There's also the upfront portion such as preparing sample sheets which again require training and LIMS tweaks; it will be interesting to see how MGI is addressing this.

Then there's the data differences. I know from someone who has compared NovaSeq versus MGI un-Cool data (yeah, I'm inventing that term) with an eye towards clinical applications that they are essentially interchangable.  In the preprint MGI claims that the CoolMPS data is even better.  For the bulk resequencing markets just total data quantity is the measure most customers care about, but some applications are pickier about read format.

MGI could just decide to gun for new players.  Not to be too cynical, but newbies are always easier to impress on pricing and getting new stuff.  Less cynically, any new installation doesn't face the same switching and training issues; a brand new start means no old constraints.  

There's also the open question of how Illumina will respond.  Illumina has big margins and could trim prices to compete, particularly with discounts on instrument purchases to keep customers in the fold and block MGI from taking new ones.  But Illumina is a public company with highly scrutinized revenue numbers, which constrains them.  Wall Street is likely to punish any strategy with significant short term impact to revenue numbers -- and will punish in the long term loss of market share.  Illumina might also have some technical tricks to unleash outside of their usual J.P. Morgan conference news release pattern, such as new flowcells or kits.  Or they could largely stand pat and rely on being the safe choice (and perhaps some more IP lawsuits).

MGI will be at AGBT in force and is even closing out the meeting with a talk claiming they can deliver $100 human genomes.  So in addition to details on CoolMPS, expect one or more updates on MGI over the next week or so.  Have an opinion on this?  Catch me at AGBT wearing my Apollo 13 hat or hit me by the usual electronic means

[2020-02-23 -- tweaked the language of T7 & NovaSeq to make clear 4 and 2 is number of flowcells running simultaneously]


Anonymous said...

The NovaSeq has 4 flowcells, not 2.

Keith Robison said...

Sorry for the confusion; I've tweaked the wording to make it clear I'm referring to T7 able to run 4 flowcells simultaneously versus only 2 slots on the NovaSeq

NovaSeq definitely has 4 stock flowcells plus many more run formats per flowcell whereas T7 is being offered with only 2 run formats total

Anonymous said...

Scar limiting read length feels a bit like FUD/propaganda. It would have to be on hell of a non local effect for a scar 300 bases away to affect what the polymerase is doing at the active site.