Thursday, December 01, 2016

Oxford Nanopore New York City Meeting, Day 1

Oxford Nanopore officially kicked off its Community Meeting in New York City today; a training session took place yesterday.  Already there have been several interesting announcements and presentations, including a new prototype sample prep gadget, a new basecaller which improves homopolymer calling, a read-both-strands approach that isn't 2D sequencing and details on multiple human genomes run on MinION. A reminder: I'm working only from tweets; I'm not at the meeting.

First, a quick note on my piece yesterday positing that PromethION is a strategic detour for ONT.  I head noted that I hadn't seen evidence that any customers had received flowcells, but apparently while I was off on my annual Thanksgiving retreat I missed.  So that's one important step forward for the PromethION program - though Clive Brown in his talk only promised that flowcells would ship soon (more below).

Clive Brown's talk had one complete surprise and another key update which had been hinted at in his "Wafer Thin Update", plus a number of checks on the progress of various prior announcements.

The surprise is a new gadget, a bead-beating sample prep add-on.  No clever name yet, but with a promised 13 minutes from sample to flowcell loading, it is yet another big promise from ONT.  If it really works on a wide range of samples (Brown apparently suggested stuffing a bug into the device's loading port), then field sequencing takes a huge step forward.

An announcement which could shift the entire competitive landscape for ONT is that a new basecaller, using "transducer" technology, is imminent.  What makes this update potentially game-changing is that this new tool, Scrappie, apparently resolves homopolymers. It apparently will call DNA modifications as well. Clearly this needs to be validated externally on a wide variety of samples, but if true it could remove the major barrier to high-quality nanopore-only de novo genome assemblies.

Brown also described a new "Follow Through" library approach.  Like the disputed 2D library prep, it would read both strands of a double-stranded template, but with no hairpin connecting them.  Going around the hairpin apparently caused the second strand to have lower quality; this is apparently eliminated with Follow Through.

Brown reviewed the "Flongle" mini-flowcell, which implements the "Crumpet" separation of the electronics (ASIC) from the wet parts.  Brown suggested Flongles would sell (presumably in bulk) for $20.  Oxford has touted this, rightly, as an entree into diagnostics.  I think such low-cost flowcells will really get educational and hobby sequencing going.

Back to PromethION.  Again, flowcells are promised very soon. Brown suggested between 5 and 20 machines have been delivered. Flowcells are specified to yield 1.5 gigabases per hour, with 60 gigabases for a full run.  That won't quite yield a 30X human genome on a single flowcell.  Flowcells are divided into four compartments, allowing four separate samples.  A whopping 5850 channels in a full flowcell.

Dan Turner from ONT gave several additional updates on sample prep.  VolTRAX instruments are starting to arrive at early access customers.  Initially supporting 1D transposase and 1D/2D ligation preps, with direct RNA prep well along (which will be released in manual format imminently).  VolTRAX is now able to perform PCR, offering the possibility of 16S identification or similar methods in a non-laboratory setting.  E.coli DNA extraction and library prep entirely on VolTRAX has been demonstrated.  Also, of great import, Turner stated that the official plural is VolTRACes (personally, I'd prefer VolTRAXen).

Oxford had announced a human genome produced on a MinION back at the American Society for Human Genetics meeting, but the data hadn't been released.  Now multiple groups have generated 20-30X genomes on MinION. A consortium has generated a dataset for NA12878, which is being made publicly available and has coverage of 99.7% of the genome.  Vladimir Benes of EMBL suggested that mapping to reference genomes is about to become pass√©; de novo assembly will be feasible for every sample.  Clive Brown gave updates on his personal sequence, for which he has performed the sample and library prep.  He's up to 150 gigabases of basecalled data, and promises to soon generate a complete dataset on a single PromethION flowcell.  Wigard Kloosterman described structural variant identification in a chromothripsis sample.

Well, it's late so I won't attempt to do justice to all of the tweet stream.  I've organized tweets into several Storify stories, grouped somewhat thematically. I apologize if the titles I've given below aren't ideal; they should give some idea of the content but don't expect more than that.

No comments: