Sunday, January 06, 2019

2019 Sequencing Tech Speculations, Part I: Illumina & MGI

Next week is the J.P. Morgan Healthcare Conference.  It's striking this year the paucity of companies in the genomics space -- Illumina on Mondayat 6:00 EST and MGI on Wednesday at 5:30 EST and Nanostring at 6:30 EST on Wednesday.  Perhaps NVIDIA will say something interesting about their forays into healthcare, such as providing the chops for real time nanopore basecalling, on Thursday at 11:30 EST. There's also some nice polls from Albert Viella on Twitter
and some more Twitter speculation

So, before all the beans are spilled, here's some of the speculations and ponderings I've been entertaining about the sequencing technology field for 2019 for Illumina and MGI.  I'll cover some of the other players in the next few days, but since Illumina is up on Monday that's the priority!

How Will Illumina Divide Its JPM Presentation?

Illumina has a broad portfolio of different businesses.  There's the venerable array business, the short read business, the acquired Pacific Biosciences long read business and then the various spin-offs in clinical and personal genomics.  One shouldn't put too much weight on this, but what fraction of presentation time will each of these components get at J.P. Morgan?

How Will Illumina Rebrand PacBio?

Illumina's acquisition of Pacific Biosciences (which has not closed, as far as I can tell) leads to all sorts of questions.  One is how quickly and in what manner Illumina will market the PacBio technology.   My own guess is that the term "Pacific Biosciences" will start disappearing before year's end but the SMRT (Single Molecule Real Time) and Sequel will last.  But I wouldn't be surprised to see a complete rebranding -- TruLong or some such thing.

How Will Illumina Integrate PacBio?

Another side of the PacBio/Illumina coin is the degree to which PacBio technology will be folded into existing products.  I will predict that an early push will put long reads firmly within the BaseSpace ecosystem -- direct dumping of data from instrument to BaseSpace, apps specific for long reads and such.  Perhaps there will be apps enabling common combined uses of the technologies -- error-correction of genome assemblies or using Iso-Seq data as the foundation for short read expression profiling.

How Aggressively Will Illumina Develop PacBio Technology?

One hope for positive effects from Illumina acquiring PacBio is that there will now be more resources to pour into further developing the platform.  Further yield improvements will probably mostly come from further increasing the number of ZMWs per flowcell.  Longer movies and shifting read length distributions help as well, but bump into the challenges of making large insert libraries.

The other direction for improvement would be to revive the idea of a benchtop instrument using the technology.  Whether such a device could be built for an attractive price is pure speculation at this point.

How Hard Can Illumina's Salesforce Push PacBio?

PacBio was ultimately an acquisition target -- selling for about 50% of its all-time high even with the offered merger premium -- because they couldn't move machines into the marketplace fast enough to drive growth and profitability.  Illumina will assuredly unleash their potent salesforce to try to get more boxes out into the world.

What would success look like?  Analysts will probably focus on numbers of machines placed, which is a good metric.  But I think some really high-profile placements could be just as big.  One example would be to place Sequels at some major biopharma companies.

A truly amazing event would be to place multiple Sequels just down the street from Starbase, at the Broad Institute.  The Broad's sequencing center has some walk-up services of MiSeq type instruments, but they have always been 99.9999999% about high throughput production centered around whatever is the current top-of-line Illumina instrument.  I've seen their PacBio RSII during a tour, but it has almost notoriously sat idle during most of its existence.

How Will Illumina Respond to MGI? And Vice Versa?

In October -- just before the PacBio acquisition announcement -- China's MGI announced a new top-of-line sequencer (MGISEQ-T7) intended to challenge the NovaSeq.  In particular, it is claimed to be able to deliver human genomes in one day versus two.  MGI also appeared to be claiming lower costs, though until the system is launched it won't be possible to get an independent assessment on that.  The T7 is also a dual quad flowcell machine, with independent operation of the two four flowcells.

Based on the GenomeWeb article when things were launched (I totally dropped the ball on covering this when it happened), the system won't be fully launched until the third quarter.  It's also unclear from that article whether the system will be launched in the U.S. Might MGI have some intellectual property worry in the U.S.?

MGI is touting that their DNB (DNA nanoball) template preparation technology generates very low index hopping, a problematic phenomenon for Illumina particularly on the newer machines using exclusion amplification on patterned flowcells. They are also offering a single-tube version of the long fragment read technology for linked read sequencing -- but have also been enabled as an instrument for 10X Genomics Chromium.

Illumina largely stole the fire the launch of the original Ion Torrent by announcing MiSeq.  Could this repeat with BGISEQ T7?  A number of the speculations I included at the head of this are for a 2x250 mode to be announced for NovaSeq; that would drive the costs down but not the time.  But perhaps Illumina has tricks for faster runs?  A trivial one would be a smaller flowcell -- ala the rapid mode on the HiSeq 2500 -- to enable one day genomes.  That would be a purely defensive move.  A robust offense would provide something to meet or beat the T7 -- offering multiple, independent flowcells.  Or perhaps a major upgrade of the NextSeq enabling faster 30X human genomes?

But some of that also depends on Illumina's perception of its customer base.  For example, if many shops are pure high throughput installations that aren't concerned with batching samples, then simply improving yield (such as with a 2x250 mode) of the NovaSeq might be sufficient

On the MGI side, perhaps we'll get some clarity on launch schedule to different markets (particularly the U.S.) 

Other Illumina Acquisition Targets?

Illumina has generally done well with acquisitions, adding useful technologies.  Some haven't worked out -- Advanced Liquid Logic is the obvious exam and Moleculo had limited impact as well -- but others such as Epicentre (for Nextera) have been huge wins.  I don't spend a lot of time thinking about this (and absolutely never make investment bets on companies I comment on), but it can be fun.  For example, 10X Genomics would be an obvious candidate -- their Chromium system can be seen as expanding the range of Illumina boxes -- and buying them would allow blocking MGI from leveraging 10X technology.  In the long term I expect the 10X linked read technology for DNA to fade in importance as true long reads get better, but they can operate off tiny amounts of DNA.  More importantly, the various single cell and specialized assays -- such as reading out T-cell or B-cell immunorepertoires -- could be very valuable to lock down.

Another class of possible Illumina targets would be the growing number of companies developing schemes for specialized library prep.  10X has been busy in this area, snagging Spatial Transcriptomics and the ATAC-Seq company, Epinomics.  There are plenty of other companies out there with library techniques for taking some sort of interesting biology or long-range information and converting them into sequencing libraries to run on Illumina.  And again, with heating competition in the short read space (as opposed to QIAGEN and Ion Torrent, which seem permanently pigeonholed to small targeted panels), acquiring companies enables blocking MGI.

Long Reads in the Clinic?

For both Illumina with the PacBio technology and Oxford Nanopore, there is a question of when long read assays will move from purely research to regular clinical practice.  After one of my prior posts, someone sent me their poster of using PacBio reads to haplotype a dominant mutation my correspondent's company is targeting.  And at the San Francisco Nanopore Community meeting there were several oncology assays -- such as identifying gene fusions in leukemias and lymphomas -- that seemed ready for prime time.  But actually getting these into the clinic means navigating the regulatory requirements (which I'll be free to admit ignorance of) and solving issues around supply chains.  There are apparently multiple CLIA labs running PacBio, for example, but I am unaware of any nanopore labs.  And if part of the value proposition for an assay is speed -- which was certainly the case for the talks in San Francisco -- then there must be an assured supply of working flowcells and reagents.

As always, comments, corrections and criticisms welcome -- pepper me in the comments (moderated, so won't appear immediately -- I pass through anything on topic) or send email to my gmail (keith.e.robison). What did I miss?  What am I completely off base on ?

[Monday morning, 5:31 EST -- fixed the T7 to quad flowcell, error caught by a reader which I thought I had already fixed]


Liang Zong said...

Hi Keith, it's reported that "MGISEQ-T7 is built with innovative quadruple flow-cell staging that allows simultaneous but independent operation of 1 to 4 flow cells in a single run", not just a dual flowcell machine.

Keith Robison said...

Thanks - thought I fixed that mistake before I hit post...

Bertrand said...

Hi Keith, thanks for a very interesting set of speculations. You mention the MGI T7 as having two flowcells, while the GenomeWeb article indicates four flowcells, is this a typo or do you have additional information? Also, are you aware of the Chinese revival of Helicos technology (a benchtop instrument called "Genocare") by a company called Direct Genomics, one of whose founders is none other than the now infamous Jiankui He?