Tuesday, November 27, 2018

Nanopore Community Meeting 2018 Preview

Okay, now that I'm done venting -- for now -- about ONT's customer service experience  (well, almost done -- they sent me the same damn letter they sent my colleague -- why were they several hours apart???) -- let's move on to the Nanopore Community Meeting.  Technically it started today with the training session, but I'm not heading out until tonight.  At the first one of these in NYC Oxford tried to avoid making any announcements, but they seem to now like having two major focus times a year sometimes supplemented with Clive Brown webinars in between.  Here are some


The smaller flowcell with Flongle attachment has been made available to a very select group of users (alas I haven't seen one in the field).  There's been a lot of tweets of good success, most recently Matt Loose:

Presumably an expansion of the number of users will be underway.  Flongle is very exciting since many applications -- targeted diagnostics, education -- can get by quite nicely on a hundred megabases.  Heck, a geezer like me can remember when that was be considered good annual output for an entire genome center!

R10 Flowcells (or, the Two Pore Whys)

oxford has been messaging that there will be announcements of a new flowcell type, R10, which will have two different pore types present.  The general idea is that the two pores will have different error profiles which will be complementary, enabling much higher accuracy consensus data.  There's a number of interesting issues arising from this.

First, as I kvetched earlier, ONT periodically has serious supply chain hiccups -- and they burped again on us very recently.  It can't help that they now have three incompatible sizes -- Flongle, MinION/GridION and PromethION -- and potentially two different pore types -- the R9.4 for 1D and R9.5 for 2D.  Now a third pore class is thrown in.  The full 3x3 grid probably won't be available -- I'm under the impression (but don't have data) that Flongle is R9.4 only.

The more interesting question is how much rewriting of tool chains will be needed to get the benefit and how will people approach this?  If the two pore types have different electrical signal properties, then one might expect that any basecaller will need to guess which pore type is running in a given channel (it would be surprising -- but not impossible -- if there were a spatial pattern to the pores).  I don't doubt that can be intuited from the signal, but its a new complication.  It also seems to be a reasonable assumption that the pore type will be added as metadata in the FASTQ headers for later tools -- polishers and modification callers and error analyzers (more on the last one soon!) -- can utilize that information.


The MinIT specialized computing device is apparently working well in the field and ONT is ramping up production to meet the demand.  MinIT uses GPUs to enable real-time basecalling; in my hands the basecalling consumes a bunch of time and power on our compute cluster.

It will also be interesting to see ONT's plans for what software to support on MinIT.  To give a very pressing example, the amazing Ryan Wick is apparently a finite resource and has understandably stopped supporting the adapter trimmer / barcode extractor Porechop so he can focus on other endeavors.  There are also newer barcode parsers out there working in electrical signal space (oh wait, look who the first author is -- Ryan Wick!) -- which is also a pressing need given the embarrassingly high number of reads missing barcode detection a number of people reported at London Calling.  And another tool that might need tuning after R10 rolls out -- signal-level barcode parsers!  Ideally all of that would be wrapped into the main base calling pipeline -- removing contaminating adapters and splitting by barcode is should all logically be part of the same operation.


Oxford apparently had the V2 version of the VolTRAX library prep instrument running at the ASHG meeting.  This device has PCR (and qPCR?) capability as well as magnetic bead handling.  Unfortunately, it doesn't appear that kits for that will be available from the start.


Many PromethION users have been reporting yields routinely in the 70-100Gb range per flowcell.  On the other hand, I've had users tell me that they can't reliably get flowcells.  That's one big plus to actually getting to a meeting -- I can try to get scuttlebutt out of people!

Cool User Stuff

I haven't looked much at the program, but it would be unfathomable that the conference won't feature great innovative work by nanopore users.  So watch this spot for updates!

What did I forget?  Tweet/DM or email me if there are topics you want to perk my ears on.

No comments: