I've thought of a number of possible contributors to Roche's decision, and for each there exists a counter-argument. Also, I want to emphasize again my high degree of respect for PacBio's technical acumen, for their pioneering long read sequencing applications and launching a slice of science fiction in the scientific marketplace. I also empathize with PacBio's employees; seeing a major partner walk away is disheartening, and seeing Wall Street knock your market cap down in one day by over 40% is certainly difficult. They've worked hard and I can't see this being their fault, but in the end the employees will feel the pain the most.
So what led Roche to walk away? I'll throw out some speculations, but I don't have much confidence in them as you can see. But perhaps my idea of rational isn't Roche's, and in any case I'm sure they have information they don't.
An obvious possibility is that Roche became disenchanted with the Sequel. While smaller and less expensive than the predecessor RS II, it is still a large machine costing multiple hundreds of thousands of dollars in a field increasingly moving to benchtop or smaller devices.. Sequel has also been slow out of the gate and not yet delivering promised throughput. This was noted in a recent, and now clearly prescient, blog post by Mick Watson. Or Roche could have decided that the market for PacBio-style reads in the clinical space was just too small to warrant the large investment still required.
But the counter-argument is that Sequel does work and it's an evolutionary update to RS II, so all the previous assays will work. PacBio still has a lead in key applications and can generate very low consensus error rates. Long reads have demonstrated utility for applications such as major histocompatibility complex haplotyping, and PacBio's lower and non-systematic error rate are still important strengths in the long read space.
PacBio hasn't developed a rich ecosystem of third-party tools, or even useful auxiliary gear from PacBio itself. Library preparation is more involved and manual than some other platforms, particularly now that Oxford Nanopore has a sub-10 minute library prep. Clinical applications requiring high levels of training and high dependence on technician skill might be a hard sell.
But if this was Roche's complaint, then the more obvious solution would be to develop the library prep automation. Doesn't seem like a deal-killer
Mick's big point in that post is that Oxford is pulling ahead of PacBio. Lower cost-of-entry, high possible yields with R9.4 chemistry, longer reads. Oxford is now launching their beta program for PromethION, which CSO Clive Brown boasts will have the throughput of 96 Sequels, but with effectively no capital cost (you spend $100K to get the machine, but it comes with a similar amount of consumables). ONT is also now launching a beta to VolTRAX, a simple, compact and inexpensive library prep device. PacBio demonstrated direct RNA sequencing, but never commercialized it; ONT is now sending out direct RNA sequencing kits to beta users.
But PacBio still has a huge lead on publications. ONT still has a serious problem with homopolymers and perhaps other systematic errors. On a good day R9.4 can heartily beat Sequel for data yield, but those days are rare. A large set of flowcells run to generate 100X coverage of a tomato cultivar showed a mean yield of 3.4Gb, not the gaudier 10Gb that ONT touts. It still appears no user has generated data on a PromethION, which I have written is a risky leap forward for Oxford. VolTRAX and direct RNA sequencing haven't yet been demonstrated to work for end users. Oxford also shows an insistence on commercializing solo, which hasn't always gone well. A user recently tweeted a photo of defective flowcells, and others have complained about shipping delays (well, I did ). Finally, Oxford isn't making money and just raised $126 million. An analysis posted in the comments of one of my scribblings suggested that Oxford has much slimmer margins than have typically existed in the industry.
The obvious culprit would be Genia, the nanopore sequencing technology which Roche purchased. Genia plus PacBio seemed like an uncomfortable hedge, as Genia could potentially supplant PacBio with long reads at higher quality on a very compact instrument. Genia has twice reported this year technical success, first with their nucleotide chemistry and then their protein engineering.
But good publications do not a commercial instrument make, and there has been no signal that a launch is imminent. Perhaps we will hear a lot more at JP Morgan or AGBT, but Genia has burned a lot of credibility in the past with premature announcements. Beta testing was supposed to start the end of 2012, then 2014 (I can't rule out they also claimed 2013, just can't find it). Many of complained that ONT announced vaporware at 2012, but they've delivered devices to customers since mid-2014. Genia is an unproven platform, whereas Sequel is ready to go.
I have no idea what specifically will happen in healthcare next year, but it is a near given that there will be great turmoil. The incoming U.S. President has announced intentions to "repeal and replace" the Affordable Care Act, has complained about drug pricing and rumored candidates for FDA Commissioner are skeptics of much of the current regulatory regime. He's also proposed major changes in corporate tax law and trade relationships. Funding levels or trends for the NIH are up in the air, as is whether Francis Collins will be a rare retainee from the current administration. The FDA is putting off emplacing regulations covering Laboratory Derived Tests (LDTs), which might both drive Sequel sales and compete with kits Roche might develop. And the market for sequencing is starting to look a bit glutted.
That's a lot of uncertainty, but really enough to scare off a diagnostics behemoth? A huge multinational which could launch in Europe and Asia while waiting for the dust to settle in the states? There might be a glut for research sequencing of human genomes and perhaps some other applications, but not much evidence that extends to clinical sequencing.
Roche Eyeing Another Sequencing Platform?
This seems like a longshot, but maybe Roche is thinking of going after a different platform and wanted anti-trust issues cleared away. But whom? Oxford shows no signs of wanting to tie up with anyone for any area of the world, let alone for everything. Roche tried for Illumina before and the market leader has not been Wall Street's darling lately and Jay Flatley is no longer CEO, but that still seems like a stretch. Why would Illumina give up independence? Would ThermoFisher be willing to part with their Ion Torrent business? The remaining companies are either small fish (e.g. SeqLL) or not yet even minnows (all the stealth-mode and semi-stealthy startups), unlikely to present a problem.
So Where From Here?
Well, both sides of my mouth and both hands are obviously worn out. It's a surprising decision, even if you think PacBio will be eventually displaced by other single molecule technologies; I don't believe anyone thinks this will happen overnight. With their financial might, Roche probably could have driven Sequel into the clinic.
Roche has an unpleasant reputation with genomics aficionados. The bought NimbleGen and their clever optical-guided microarray synthesis platform, but didn't develop the technology significantly and ultimately kept little but the name. Roche bought 454 and then proceeded to demonstrate that being first mover brings no guarantee of success. Again, a failure to keep up with competitors by pushing the platform to greater densities of clusters, plus a failure to at least thwart Illumina from acquiring key components of a complete platform. For example, Illumina bought EpiCentre for their Nextera library prep technology, and quickly shut down the kits for the 454 platform. If Roche had beaten them to the punch, then at least they would have had some advantages over Illumina.
PacBio has indicated they will now commercialize Sequel into the clinical space. Certainly there will be a lot of intense strategy sessions and quite likely some high level hires with experience in the clinical space. PacBio has indicated that they will focus on LDTs, which means be very tied to the uncertainty around the FDA's regulation in this space. Still, supporting academic investigators or entrepreneurs who develop diagnostics is an easier road then doing it yourself, so the strategy makes good sense for a small company that isn't yet profitable.
I suspect the mood was tense at PacBio today, but a good team such as they have will reboot and get cracking. Not only do they have the normal incentive of wanting to succeed, but now they have the added impetus of wanting to make Roche's management look like fools.