Wednesday, May 25, 2016

London Calling Preview

ON Thursday and Friday this week Oxford Nanopore will be holding their second annual London Calling meeting.  I successfully defended my schedule this year, so I'll be on the ground there. If you follow me on Twitter and don't want to be buried in nanopore tweets, mute the hashtag #nanoporeconf (a rather large hashtag for talking about nano stuff!)  LC is OxNano's premier event, so what might we see from the company?


Oxford announced a long list of platform improvements at  last year's Calling, to be released across the course of the year.  Unfortunately, few of these hit wide distribution, and updates at the U.S. User's Meeting and then a March "No Thank's We've Already Got One" webinar adjusted schedules, but didn't trigger widespread releases.  With that in mind, here is what I'm most eager to hear from Oxford itself or as major announcements from users.  A lot of these are interconnected, but I've tried to disentangle them

The 1D transposon-based library kit promises 20 minute preps from purified DNA using only minimal skills.  Nick Loman and Matt Loose used this kit to prepare a library in their AGBT hotel room.  Those who have used it have tweeted great happiness with the kit and its ease of use, not to mention a read length distribution with a very long tail of mongo reads.  However, that is a select group of testers -- the kit is not in general circulation yet.  Clive Brown has indicated on Twitter that the kit has had some stability issues in shipment, but also holding this up is the R9 delay.

R9 is the next pore chemistry from Oxford, which they announced at the "No Thanks" webinar.  Faster with higher accuracy, plus a completely clear intellectual property position, the best of all worlds.  And indeed, those who have used it have been raving at the high accuracy, high yield and length of the reads. But again, R9 hasn't hit the general community.

PromethION is Oxford's answer for population-scale sequencing and other megasequencing projects.  Oxford indicated by Twitter shipment of the first PromethION on April 21st to an undisclosed site.  In addition to trying to glean from the grapevine the location of that first machine, I'll be very interested what data has been generated.   I had predicted back in March that if a device was received early enough and if it worked well, the lucky lab would try to generate a human genome sequence from it.  Did such a dataset get generated in the last month?

For PromethION's to be economical, Oxford was planning to split the current consumable into two bits -- the expensive electronics (ASIC) would be reused, while relatively cheap plastic on the wet side would be disposable.  This "crumpet chip" would also enable much lower overall costs for MinION users, as well as the possibility of "pay-as-you-go" and other pricing models akin to how cell phone service is sold. Such low capital cost (or no capital cost) plans could be very valuable for the DIYBio, hobby and educational markets, as well as a boon to small biotechs and to academics with limited funds.

A key boost for productivity would be increasing the number of pores per flowcell.  At last year's LC, ONT suggested that boosts might be seen this spring.  Will revised guidance be given?

An issue which ONT surfaced at the "No Thanks" call is a problem with declining numbers of active pores during the running of flowcells.  This degraded performance steals perhaps as much as 50% of the theoretical yield of a flowcell as well as making reuse of flowcells less attractive.

VolTRAX is ONT's microfluidic vision for sample prep, which demonstrated the moving around of colored water back in New York.  Seeing a VolTRAX do something more would be an important step.

Direct RNA sequencing is an exciting possibility broached last year at London Calling, as it would eliminate the cost and errors (such as chimeras from strand-switching)  associated with reverse transcriptase.  ONT had indicated one sticking point in finding a good motor; tuning the base-calling model would also be necessary.

So batten down your browsers -- it promises to be an exciting few days of red-hot Tweeting.


7 comments:

Geneticist from the East said...

Good to see that someone had their hands on R9 already. Are there any R9 preprints to read and/or R9 data to download?

Keith Robison said...

R9 E.coli data released today

Geneticist from the East said...

Wow! Thanks a lot for your link. 179GB should take a while to download though.

Do you know if this data works with poRe or poretools?

Keith Robison said...

I believe you'll have no problem - Nick was part of the generation so poretools should be fine & I believe Mick has had access to R9 so poRe should be good - but I'll try to catch both of them. I think for those two the main issue would be FAST5 format changes; signal-level tools may need new models.

Nick Loman said...

FASTA/FASTQ extraction doesn't seem to be affected by new changes. Nanopolish needs updating to support the R9 model because there is no longer HMM parameters or scale/shift values in the FAST5. We've made those changes internally but they are not yet ready for release.

Geneticist from the East said...

I am still in the process of downloading the tar file. But I untar a few fast5s and looked into them. I noticed that the fast5s are formatted as metrichor1.16 which has /Analyses/Basecall_1D_000 as oppose to 2D_000 according to poretools' Fast5File.py. I suppose that means these fast5 files are basecalled by classic HMM but not the latest RNN method, right?

https://www.nanoporetech.com/news/press-releases-and-announcements/view/49

This recent press release seems to imply that the new Rapid Sequencing Kit only produces 1D reads. Does that imply the tar file I am downloading is using the old prep kit that has hairpin adapters and therefore has both template and complement and produces 2D reads?

Nick Loman said...

Those R9 files are called by the RNN.

Yes, this library is made with the standard 2D sequencing kit rather than the rapid sequencing kit. You can tell the rapid sequencing kit because the read length distribution should look log-normal.