Wednesday, May 20, 2015

London Calling Wrap-Up

The second, and final, day of Oxford Nanopore's London Calling conference concluded last Friday -- and I'm behind on writing it up.  Some of that was due to travel (and the wrong power supply going on the trip) and post-trip exhaustion, but failing to finish this last night was pure slacking. That route was shut down when one reader asked when I'd get things done.  Anyway, I again organized the activity into a storify story as I did for the first day of the conference. I'm going to go into less detail on individual presentations below and instead engage in the vice of far-ranging speculation.



Before I jump into some broad comments, there is an outlier that doesn't fit well -- proteins!  Michael Mayer from U. Michigan capped the conference with a talk on analyzing proteins with nanopores.  Unfortunately, there wasn't a lot of detail in the tweet stream on this, but the methodology is described as "5D" -- measuring 5 properties of proteins, which may be along the lines of size and general shape and charge and such.  Very early days, but definitely a lot of excitement.

It is already clear that the Nanopore community can be subgrouped into several constituencies, with differing demands. Already some have tweeted that they want to see less of subgroup X talks because they care about subgroup Y.  I'll call these four subgroups: Fast, Field, Long and Lots. For each, the name suggests the attribute that is most treasured by that group. Some of these definitely track together, and each is likely to bump elbows with different facets of the existing sequencing platform market. Many improvements in the platform will benefit all of these groups, but there will be definitely some that trigger one faction to scream "want, want, WANT!" while others shrug a "meh". 

The Fast crowd wants speed over everything else; their scientific questions are time-sensitive, such as diagnosing infectious agents or monitoring the environment in nearly real time.  Time to result is most important here, with other qualities less valued.  For example, only a small number of reads of relatively modest accuracy and length may be sufficient to flag an antibiotic resistance island.  Fast may commonly associate with Field; often Field applications value speed as well.  But many Fast applications will be run in hospitals or other settings with reliable high-bandwidth internet access, which is a differentiator from the pure Field group. Fast applications are often going to have PCR as a front end, to focus the analysis on what matters. Several existing platforms, such as MiSeq, Ion Torrent, 454 and even PacBio, have previously been applied to Fast challenges, but this is a market where Oxford is likely to gain traction quickly, for those applications where the sequence quality is sufficient.

It's worth noting that even within Fast there may end up being a division, between those for whom a day from sample to result is adequate, and those for whom an hour is too slow. For example, Ron Ammar spoke on day 2 on typing pharmacogenomic variants and there was the previous day's talk on sepsis as examples of uses with very tight deadlines.  For sepsis minutes can make a clinical difference, but for any test that might inform treatment decisions time is precious.  Clive Brown's statement that an ultra-rapid but 1D-only prep is coming might only appeal to the ultra-rapid crowd: it's a small subset of problems which 60%-ish identity reads can address (though one of the breakout sessions apparently had some optimism that 85+% might be attainable from 1D)

Field partisans want to take the instruments anywhere in the world (or even out-of-this-world; a MinION has apparently made it to the International Space Station).  Portability -- not just of the sequencer but the entire operation -- are paramount.  At least two speakers working on Ebola commented that local temperatures in West Africa interfered with PCR operation, though this could apparently be solved in one case by using a metal plate as a heatsink.  As suggested above, the current requirement for high bandwidth internet to enable cloud calling a serious issue for the Field group -- though Josh Quick apparently was able to buy up a set of SIM cards and successfully sequence over a 3G connection. The Field crowd will put a lot of pressure on the upstream tool chain; your standard thermocycler is not something you want to lug around in a backpack! Clever instrumentation -- or switching to isothermal assays -- will be popular in  this crowd.  Talks on day 2 in this space not only included two on Ebola, but also tracking illegal timber trafficking, river water microbiomes, wildlife profiling and food safety. The only platform that's seen much field use as far as I can tell has been Ion Torrent PGM, which is still bulky but at least has no optics.  Another market where Oxford should find fertile ground.

The Long posse wants to assemble genomes, resolve haplotypes and sift through metagenomes. They want really long; ideally tens or hundreds of kilobasepairs.  It's worth noting here that many of the Fast and Field applications described in London require (or benefit from) reads that are still pretty long; one of the Ebola assays used 8Kb PCR amplicons!  In addition to a number of talks at the conference, a new pre-print from a MAP group discusses obtaining 100kb reads (by ONT), and also generated some interesting preliminary evidence that G4 DNA structures (a 4-stranded structure that forms with certain G-rich motifs) might impede passage through the pores.  Oxford has generally been very confident that their platform is immune to secondary structure, so this would be an interesting suprise -- though hardly a disaster.  Long is obviously impinging on PacBio as I scribbled recently, but if the read lengths can be pushed further Oxford may be trouble for the optical mapping companies as well.

Finally, perhaps the least represented at the conference was the Lots group, which has an insatiable appetite for data (which currently only HiSeqs can attempt to satiate).  Of course, every group will want lots of data, but some applications truly demand it.  The conference didn't seem to to have many (any?) talks on these lines, perhaps because the current throughput of MinION is considered too small. This crowd is likely to want to interface these devices with automation -- the MinION is certainly NOT automation-friendly (among other things, there is a little cover over the sample port which takes some dexterity to work), which is unfortunate since it must be fed with fresh library and fuel every 6 hours.  PromethION will likely ignite interest from the Lots crowd for complex genome sequencing, metagenomes, RNA-Seq, various counting assays, etc.  

The VolTRAX sample preparation clip-on is very interesting, though availability (summer?) and cost (??) are fuzzy.  An interesting question is how much of the workflow will be moved off onto VolTRAX; presumably the initial input will be amplicons or fragmented DNA, but as cDNA/RNA protocols come on line it will be interesting to see how quickly they are adapted for VolTRAX.

Some sequencing platforms, most notably Illumina, have engendered ecosystems of supporting technologies from other companies.  Because so much of Oxford's sample preparation involves proprietary components, don't expect third party library kits ever. If other companies are to jump on Oxford's bandwagon, upstream sample preparation is the opportunity.  For amplicon Field projects, lugging a conventional thermocycler around (and finding an outlet to plug it to) will probably be a burden which might finally drive microfluidic PCR devices into wide use.  Of course, the ideal instrument would combine PCR, library preparation and sequencing all in a single self-contained unit.

The first 11 months of MinIONs in the wild have seen both amazing hints of fulfilling the promises of 2012 as well as mundane struggles with shipping and customs and bubble "snail trails".  ONT has announced a dizzying list of platform upgrades, which the growing MAP community (and MAP is apparently the way ONT is launching the product) awaits eagerly.    The London Calling conference has stoked enthusiasm and excitement; now ONT must deliver.

5 comments:

Unknown said...

Hi Keith, many thanks for writing this up. Any word on when the MAP program would end and the MinION would be commercially available?

Also from AGBT in February it was clear that of the two customer speakers, there wasn't any metrics around what accuracy they were seeing. Could you comment on that?

Thanks and will be sure to reach out to you the next time I'm in the Boston area.

All the best - Dale

Keith Robison said...

Dale:

The word on MAP is apparently this is the roll-out for now; ONT will simply keep issuing rounds of invitations to join. ONT has an online store up-and-running for MAP participants to order devices and consumables beyond what comes with the MAP entry fee.

For accuracy, that's still changing rapidly & the informatics and also with the application. For example,
Loman, Quick & Simpson achieved 99.4% accuracy in an E.coli assembly using only nanopore data (alas, 0.006 x 4.6M is a lot of errors) using self-correction and a polishing stage (98.4% accuracy before polishing). Urban et al say their 102Kb 2D read had a Q value of 8.74, or 87% accuracy. Szalay & Golovchenko report assembling lambdaphage to 99% accuracy.

Glad to hear you found the write-up useful

Keith

Rick said...

Yes, thank you for the write-up.

It seems to me that Field has to be the primary focus with Fastest (in less than a day) a secondary application. Fast (in a day) can be taken care of by other technologies as can Long and, of course, Lots.

Keith Robison said...

For Long, ONT will probably be superior to PacBio in terms of overall read length and distribution, so for de novo assembly & scaffolding might quickly become the dominant platform

Nick Loman said...

Keith, fantastic summary, I think you've more or less nailed it, amazing achievement considering you weren't even there.

A couple of minor corrections for the record::
- MinION isn't on the ISS yet, there was a live link with NASA who are planning on taking it up there, but waiting on the various bureaucratic hurdles right now.

- 4 & 6kb amplicons were achievable from Ebola RT-PCR, although these were hard to get. We went for a 2kb scheme in the end.