Tuesday, January 15, 2019

2019 Sequencing Tech Speculations: Will We Actually See New Entrants?

An astute reader caught a sentence fragment about MGI in last night's Illumina JPM roundup -- the unfortunate evidence of a a mental battle over whether to put any further comments on MGI in an Illumina-centric post. So now I'll sweep that bit into a general post about not-Illumina (and not-Oxford, that will go in yet another).
At J.P. Morgan MGI announced that they have hit 1000 installed instruments worldwide across their fleet. They also announced the first site to beta test their new top-of-line machine, the BGISEQ-T7 (which, alas, comes in a functional rectangular prism case rather than being styled as a big stalked icosahedron like its certain namesake).  That customer is China's largest provider of personal genome sequencing. Also announced is a marketing cross-partnership with Paragon Genomics -- Paragon will distribute MGI's library prep robot in the U.S. and MGI will distribute Paragon's gene panel kits. 

MGI is an evolution of the Complete Genomics technology, but apparently now with polymerase chemistry rather than the ligation chemistry for the actual signal generation (Anthony Beckhouse down under is MGI's most prominent user and the best source of hard information on the platform). If we count MGI as dating back to Complete Genomics days, then the last really important new sequencing company to enter the market was Oxford Nanopore, and they first started shipping in their early access program back in 2014. That was also the year that Roche acquired the Genia technology, which still hasn't seen daylight -- and rumor is that when it does it will be targeting only the diagnostics market and not research.  Somewhere in there QIAGEN launched their GeneReader, but they too have disdained research labs and focused on providing an in house sample-to-results vertical pipeline -- and they've never shown any interest in blog-based publicity, starting with turning down an overture at AGBT.

But, as Illumina's Francis deSouza noted in the JPM Q&A, there's still been plenty of venture capital interest in new sequencing technologies -- and there's activity by established players.  Nava Whiteford on his 41J blog has a nice summary of financial activity in the space (he also digs through Glassdoor to find which outfits have disgruntled employees as their major product). Small companies getting that money are very difficult to track.  Since such companies control their news and don't have customers to regularly update, it can be hard to tell if they are even still in business.  Many don't bother with websites, so lack of that isn't necessarily a tell. 

For example, shopping and restaurant complex I find myself at periodically is near ZS Genetics' last known location and I think their name is still on the signage, but that could just reflect the low rent, low demand district out in Wakefield.  ZS was trying to develop uber-long reads using transmission electron microscopy, which sounds attractive except for the hurdle of TEM is a monochrome technology so you must find four heavy atom labels which will give sufficient greyscale.

Last summer OmniOme popped up for a moment to announce $60 million in venture funding and some broad, vague promises of an amazing platform -- and since then has been quiet again.  Roche announced a $15M investment in nanopore platform Stratos Genomics, saying it synergized with the still submerged Genia technology.  Agilent bought out LaserGen, which was working on a terminator chemistry.  Agilent actually launching a sequencer would be an interesting change and Agilent would certainly have the financial muscle to be a serious player (Agilent is about half the market cap of Illumina)

The boldest statement yet came from J.P. Morgan, as Genapsys -- which had a series C round a year ago -- announced that they plan to start beta testing later this year (covered nicely in this GenomeWeb piece).  Genapsys had made a bold set of claims back at 2014 AGBT in a Steve Jobs / Jonathon Rothberg style high flash presentation.  The technology outlined used polymerases and flows of native nucleotides (ala Ion Torrent or 454) with electronic detection.  Read lengths were promised to approach a kilobase and the homopolymer issue was allegedly solved.  As noted below, I hope to soon provide a piece dedicated to guestimating the Genapsys GENIUS based on some informative patents.

The website images look about how I remember them -- a small benchtop instrument and flowcells that look a lot like Ion Torrent flowcells.  The GW article quotes an instrument price of $10K, run costs -- not including library prep -- of $300 and one day turnaround.  Error rates are claimed to be very low -- in the phred 30s -- based on Genome In A Bottle samples and this data the company says they will publish (clock's ticking folks...).  Total yield wasn't described, but chips with 1M and 16M sensors are apparently ready and a 144M sensor chip in development -- and one of these can apparently run exomes.  No discussion of loading densities -- perhaps we should just assume Poisson.  Initial placements will be at Stanford (where the inventors came from) and HudsonAlpha -- so Shawn Levy will beat Chris Mason for a first crack on this system.

Another company making some splashes is Roswell Biotechnologies, which snagged a round of funding.  Roswell claims their technology can get to sub-$100 human genomes. They also got a rosy GEN piece.  I need to do some reading on what is publicly available -- they suspend individual polymerases over electronic sensors by first tethering a scaffold DNA molecule like a tightrope over the sensor and then tacking the polymerase onto that scaffold.

So how much of this will actually break out in 2019?  Genapsys is staking the boldest position, promising instruments in the field and data in publications.  Of course, anyone who is serious about getting data out can easily post to BioRXiv, so please no excuses about anything stuck in the review process.  Proof-of-principle publications (or again, pre-prints) would be nice but no guarantees -- Genia had two nice publications in 2016 (which I covered: chemistry and polymerase engineering) yet still no instrument.  And of course there's always your friendly neighborhood blogger, for the price of an hour or so on the phone (or over coffee, if you find yourself in the Boston area).

There's still great opportunity in the sequencing space, but the existing players do keep raising the bar.  I'm not the green eyeshade type at heart, but both Illumina and Oxford Nanopore should be well under the $1K genome and the PacBio technology isn't far behind.  But there's all sorts of details left behind there -- Oxford needing to improve accuracy, PacBio being even more expensive in the highly accurate CCS mode (which still leaves some homopolymer issues), high capital costs and low portability.  And in general, unless your instrument is truly revolutionary in throughput, any buy-in price above the Sequel ($300K) is probably not going to fly well -- which means big instrument technologies such as TEM are probably nearly non-starters.

At the low end, MinION and iSeq make entering the market a difficult game of limbo.  Genapsys' system looks like they will try to outperform iSeq on throughput while coming in at a lower capital cost -- while beating nanopore on accuracy.  Oxford has demonstrated that excitement exists for instruments that can go into the field or low resource environments -- but can you really beat the MinION in that space?

And I hope all of these companies have look at past history and learned the lesson that bioinformatics matters and they should start getting datasets out and getting key developers on board.  Now, if you just have short reads in FASTQ files that look a lot like Illumina data, there isn't much to do.  But both PacBio and Oxford couldn't reach their potential without dedicated toolchains, and the same applies for linked reads such as 10X Genomics.  So if your platform's data has any quirks relative to the established players -- and that particularly includes any clever new file formats that theoretically serve your platform better than the plethora of existing ones -- think this out carefully or you may hobble your forward progress in the marketplace.  Looking at one of their patents (which I'll try to write up soon), Genapsys may well need something beyond FASTQ to represent some important details.

Speaking of patents, that is a minefield which any of these new entrants must cross -- and may want to team up with a deep-pocketed big company for assistance.  All aspects of the process -- template prep, sample prep, sensor designs, enzymology, etc. -- are subject to thickets of patent filings.  One comment on the Illumina piece mentioned that some early sequencing-by-synthesis patents should be expiring soon, which seems about right.  Indeed, I think some of the bridge PCR patents expired a year or so ago.  And the patent threat can't be taken lightly: Oxford Nanopore was apparently able to evade serious damage from Illumina's and PacBio's jabs -- abandoning the 2D format and replacing it with 1D^2 -- but QIAGEN was kept out of the U.S. market for a significant time due to an Illumina lawsuit.

Anyone who knows me will be unsurprised that I would loooooove some new entrants -- new ideas to talk about, new niches opened up for sequencing and perhaps some sharp competition to drive prices down. Let's hope 2019 ends the all-too-long drought of completely new sequencing platforms!

4 comments:

Bertrand said...

Hi Keith, just one addition to your survey: Direct Genomics in China (directgenomics.com) is marketing a single molecule sequencer that uses Helicos technology. They call it Genocare and it is clearly aimed at the clinical qequencing market. I don't know if they have actually shipped instruments (they quote the release of a beta version in June 2017). Interestingly, one of the founders of this company (in 2012)is Jiankui He, the now infamous originator of "CRISPR babies"...

Daniel Swan said...

No one seems to mention Base4 either:

http://www.base4.co.uk/

Still in quite a stealthy mode, microdroplet sequencing approach. Closed a financing round at the end of 2017, patents granted around the technology in the USA.

Unknown said...

Thanks for the write up! Made more sense once I wasn't reading MGI as McDonnell Genome Institute...

Dale Yuzuki said...

Hey Keith, thanks for the nice writeup!

You neglected to mention GnuBio, acquired in 2014 and shuttered in 2017 - here's the GenomeWeb announcement https://www.genomeweb.com/sequencing/bio-rad-shutters-former-gnubio-facility-part-digital-biology-rd-consolidation#.XEIwEM17nRa

A few others in San Diego I am aware of but may not want the publicity at this time - I can think of three of them not mentioned in your piece. 2019 should be an exciting year!

Dale