PacBio Sequel: Smaller Box, Bigger Bang

Boy, am I regretting taking a vacation from online due to being engrossed in A Canticle for Leibowitz.  Between last night and this morning, my neglect of my Twitter feed meant a colleague tipped me to the new PacBio machine with "what's this Sequel I keep hearing about from PacBio".  So a lot of folks had a huge jump and covered it pretty well, including Keith Bradnam, Mick Watson, and James Hadfield.  Long rumored, the new instrument costs about half as much (but that's still $350K), takes up much less floor space (and doesn't need any reinforced floors) yet the new flowcells deliver about 6 7 times as many reads than the older ones.  WOW!

PacBio has been a roll, launching new applications (such as Iso-Seq for mRNA) and showing the ability to assemble large diploid (e.g. human) genomes to near completion with the help of optical mapping from BioNano.  PacBio was nice enough to send me that paper (as was a kind reader), and it is worth putting that in some perspective. Thirty years ago when I was finishing grad school, sequencing the human genome the first time meant true sequencing factories, with huge banks of machines prepping DNA and running sequences, and even more factories generating the necessary physical maps to tie it all together.  BACs, radiation hybrid maps, etc, etc, all to get that first draft.  The sequencing component kept shrinking, but essentially every human genome was really a matter of mapping new data onto that old scaffold which took more than a decade to build.  Now, with the equivalent of a small room of equipment, PacBio can deliver a de novo assembly superior to that original human draft, while also delivering structural variation from an individual.  That's real progress that the notorious Moore's Law comparison doesn't really deliver properly.

With the new box, which will arrive late this year at select U.S. sites and sometime next year to the rest of the world, it is estimated that a modest coverage human genome might cost only $3K or so in reagents.  So this will still be more expensive than an Illumina X10 genome by a factor of at least 3X -- but potentially  delivering much more resolution of structural variants.  Expect also that this higher throughput / $ will mean many more metagenomic studies with PacBio.  The Sequel is a product of PacBio's collaboration with Roche Diagnostics, so there will likely be a push for putting the new boxes in a clinical setting (though right now it is research-only).

It's also interesting to think that 6X performance making PacBio even more attractive for rapid applications. I'm not sure how long library prep takes, but in contrast to Illumina that is much about fragment sizing (typically on the BluePippin) and not with PCR (as with Illumina).  For sequencing a bacterial genome, is it plausible to go from culture to high quality finished genome in a working day?  Or long PCR products for clinical applications?

Looming over PacBio is the spectre of Oxford Nanopore, which is the other single molecule long read technology in the field.  ONT's PromethION is potentially direct competition for Sequel, but has yet to be seen in the field.  If ONT is successful in rolling out their MinION improvements announced in May, even the pocket-sized MinION will be a significant competitive threat.  But similarly, none of those improvements have seen daylight yet, though most aren't expected until next year.  The battle will not only be in terms of upfront and running costs, but also on data, as while the number of de novo Oxford-only bacterial genomes is growing (including E.coli, Bacteroides & Bordetella), the final quality has not been as high and the question of whether errors are completely random has not been well addressed.  

Some commentators (such as Mick) see PacBio now as a serious threat to Illumina and apparently Illumina's stock took a 1911% dive today,  Illumina might finally have some competition, but they remain 90X the market cap of PacBio (as of tonight) with a very broad array of activities in genomics.  Even if 2016 sees  a lot of Sequel boxes in the field, there will be a lot of labs sticking to their Illumina boxes, either because its familiar or because they are working in spaces where numbers of reads are critical (such as RNA-Seq or ChIP-Seq) or where long reads just aren't a reality given the sample (FFPE for cancer samples).  Illumina announced virtually nothing at their recent Boston User Group meeting (new NextSeq chemistry with better spectral separation of the signals), but perhaps they will now be spurred -- or perhaps Flatley and company had some surprises already queued up.  There was a recent meeting item about 2x400 on the MiSeq (with some community response that first they need to fix the 2x300 data quality problem which has smoldered all year).

At the beginning of the year I made a semi-prediction that this year sequencing platforms would see evolution, not revolution.  I've been quite pleased that there have been a string of announcements that test that claim, mostly in the long range information space: 10X, Dovetail and now a big jump in PacBio throughput.  On the other hand, I'd score Revolocity and Illumina's lineup tweaks as pure evolution; neither gives me a truly new option on the sequencing technology smorgasbord.  Perhaps BGI's desktop sequencer will impress me more, but I'm not hopeful.  Perhaps the last three months of the year will smash my January pessimism -- Oxford fast mode or VolTRAX.  Or a bolt out of the blue?  One can always hope.

(strikeouts -- corrections based on comments or messages; I lazily took the ILMN stock change from that other genomics blogging Keith & he's pointed out the error)