This is a pivotal year for Oxford Nanopore as a company since they have finally succumbed to temptation and will hold an initial public offering later this year -- or float the company as I believe the English would have it. For employees that will mean liquidity of their equity positions, but for management it will mean additional scrutiny and pressure to generate revenues and profits. ONT has amazing technology with many amazing applications which are always the treat to see at London Calling and the Community Meeting. But being public may expose them to something along the lines of a catchphrase a Warp Drive board member was said to have -- when presented with our dazzling scientific work he would say "That's nice, but where are the drugs?". When I talk to financial types about nanopore sequencing, they often have a homologous response -- how is really cool application X going to turn into lots of revenue at scale Y? And what is the company doing to enable that, given that their chosen mode is to supply the tools and let the developers develop the applications.
The IPO, and a recent pre-IPO funding, means coppers to apply to expand their markets. How will they spend it? Improving distribution networks in the developing world to leverage the usage during the pandemic? Generating new LamPORE assays to address diseases other than COVID and perhaps for other than humans? Trying to enable more companies in a broader ecosystem of upstream prep kits -- only a few other companies (e.g. Circulomics, Revolugen, NEB) currently make nanopore-targeted products? New initiatives to enable companies wanting to develop high value nanopore applications? Expanded company labs for collaborating with customers to format a given procedure onto nanopore?
High Density Devices
At the Community Meeting the company presented progress on their voltage(instead of current) sensing schemes for packing many more channels onto a single chip. These promise to amp up (well, volt up) the data volumes substantially. If I recall correctly, in December they were getting useful signal and able to decode it; how much farther has this moved and will devices actually start being available for testing late this year? And what hurdles will the huge data volumes present in terms of extracting data from the devices or basecalling -- how big and pricey a GPU box will these devices require?
Q20 & Bonito & Pair Decoding
After announcing a number of advances in output sequence quality after the community meeting, Nanopore started teasing with data on a Q20 chemistry -- 1% error rate. Details have been limited, but this apparently involves a slower motor protein and a special version of the Bonito basecaller. Several alpha users have posted glowing reviews, saying they have assembled prokaryotes only with nanopore data and gotten apparently perfect genomes.
So what are the details? How is this accomplished? And what is the catch -- how much less data does this mode generate? And what quirks does it still retain -- what systematic errors? How does it perform on really difficult problems? For example, I've recently sequenced some filamentous fungi -- or should I say filaaaaaaaaaaaaaaaaamentttttttttttous funggggggggggi -- they really do have a propensity for extremely long homopolymeric repeats. If I throw Q20 at these, can it really accurately resolve homopolymers of 20 or 30 nucleotides? With what depth and how much polishing? To what degree can filtering the FASTQs get higher quality -- the Guppy quality model I've found to be quite good so perhaps one can pick only the best reads. But again, how much would that attrit the data?
Speaking of Bonito, it was supposed to show up in MinKNOW a few months ago. Any updates? I'm a bit lazy about re-calling my data on GPUs and would really like Bonito out of the gate. And is Bonito generating Q-scores yet?
There's also the concept of pair decoding and amplification-based schemes for combining multiple passes on homologous strands or copies of the same strand. When are these going to be kits?
One issue with every long read platform is you tend to need a lot of DNA upfront. An interesting calculation, which I would assign to my students if I were a professor, would be to actually calculate how many DNA molecules are called for in a library prep, how many one loads on a flowcell -- and how many are actually sequenced. The latter I suspect is rounding error on either of the others. ONT has had a number of stepwise improvements in this over the years -- tethering the library molecules so they must only perform a 2D search in the membrane rather than a 3D search in solution to find a pore, reducing the propensity of the tether to stick to all the plastic in the device, etc. But still the amounts required up front are difficult for some applications unless one amplifies, and amplification takes away some of the advantages -- such as methylation detection -- of the platform.
Nuclease On Board
Biology is a science of complexity and exceptions, and nanopore sequencing is no different. Particularly on long DNAs or certain species -- such as chicken -- the ability of the helicase motor proteins to drive DNA through the pores can be impeded by secondary structures. Reversing the driving voltage has been a nice automated solution from early on, but some molecules become so tangled as to be unable to be extracted no matter which way you try to force the physics -- a hairball formed on each side of the membrane. So the nuclease flush was invented, which restores yields but necessarily sacrifices the library. The ideal solution would be a nuclease on the trans side of the device to destroy library molecules after they are sequenced; this would guarantee no hairball on one side. But of course one can't let popped pores or anything else leak such a nuclease to the cis side. If I recall correctly, there are complications with the existing setup for trying to do this whereas it may be more feasible with the high density schemes. Still, ONT are the wizards who have solved other knotty problems.
New Consumable and Hardware Formats
I think it was two years ago at the last in person London Calling that ONT got me excited with Plongle, a 96-well format device targeting high throughput people like myself in applications such as drug screening. Any updates?
In a different direction, the GridION X5 has been around a long time -- will we ever see a higher plex device using MinION class flowcells? Or better yet, a GridFlongle with lots of flongle class flowcells? I'm thinking here as always about biotech startups where physical real estate is often under crazy constraints - GridION is a small box but not a tiny one and having more devices on it I think would have a market (yes, you can plug MinIONs into the USBs)
Then there's always the quirkier hardware initiatives -- sequencing on a cell phone with SmidgION or prepping Clive Brown's genome on stage with a simple device
VolTRAX is a concept that always seems to smolder, never catch fire. It's intriguing all the things one might do with it if all it's various bells and whistles were successfully enabled - on board PCR or on board extraction with quantitation and so forth. Or the really out there idea of performing the sequencing on VolTRAX device itself
What Did I Miss?
What didn't I think of? Probably loads, given my lack of prep -- and ONT's propensity for pulling rabbits from hats. My forehead slapping will commence in just a few hours.