Monday, June 29, 2020

Virtual London Calling, Veritably Late, Part II: Platform Development

Last time, I covered Oxford Nanopore LamPORE COVID-19 detection scheme.  London Calling was over a week ago, so the chance to scribble before its all old news is rapidly shrinking.  As noted yesterday, Clive Brown didn't speak here but instead will broadcast at some future date; it was left to his top technical lieutenants to cover the developments in the platform which have happened since the Community Meeting in New York back in early December.  I've tried to hit the highlights here, but don't claim to be comprehensive.

PromethION Fixes & R10.3

PromethION flowcells have apparently suffered from reliability problems; ONT is now promising new flowcells which fix this.

R10.3 is finally arriving for both PromethION and Flongle

More Data Per Flowcell

Next month, ONT will roll out "Kit 10" of their ligation protocol.  This features the new motor biochemistry which greatly reduces the amount of wasteful idling by motors waiting their turn for a pore.  Short insert libraries are particularly subject to this annoying tax, as they have many more adapted (and therefore motor-bound ) fragments.

In the Fall, Oxford will be offering for Early Access two different interesting improvements on the overall chemistry, one enhancing sensitivity and the other running time.

For sensitivity, the key is in a new tether chemistry which specifically binds to the membranes and not all over the flowcell.  ONT has spoken about this issue in multiple prior updates. The goal of the "trans-tether" is to get a 200 fold sensitivity boost, since the membrane is only 1/200th of the surface area inside  a flowcell.  So far they've gotten 10 fold, but if they get close to their goal they are promising that libraries of only picograms or perhaps even femtograms of input would be possible.

The longer life flowcell feels like a new thread in the complex ONT idea weave.  There are two strategies for pursuing this, both around the electrochemical mediator which is depleted during a sequencing run.  One angle is to tune the balance of components of this mediator to give longer running time; this apparently doubles the running time of pores.  The other is to re-engineer the chip details near the pore so that the wells containing mediator simply have larger volume and hence more mediator.  ONT reports runs of 40 hours on a single mux using this; compare to the 12 hours of the original MinION flowcells.  

A Glimpse This Year of Next Generation Flowcells?

ONT announced they hope to release flowcells by year's end based on a new spacing (aka pitch) enabled by changing from current sensing to voltage sensing.  Current devices have 200 micron pitch; these are aiming for 20 micron.  ONT boasts they have gotten squiggles from these. These will also rely on new Application Specific Integrated Circuit designs which are small, cheap and disposable.

Both the SmidgION cell phone accessory and the 96-well Plongle devices rely on the new ASICs.  These got only a mention on a slide without any promises of when they would appear.

More Barcodes

I've been one to grouse that ONT didn't have enough barcodes for my tastes.  I like lots of samples, both back at Warp and now at Ginkgo, and having only a dozen or so didn't suit me.  This month ONT expects to deliver 96 native barcodes for ligation, followed with 48 cDNA and 96 rapid barcodes around the end of August.


In my opinion the VolTRAX microfluidic library prep device has never really taken off; you don't see hordes of people tweeting about it.  ONT is continuing to expand the device's capabilities, so perhaps a critical threshold for wide acceptance will be crossed with one of these.  Around October they will be releasing VolTRAX capable of PCR on device, which will enable a version of the ARTIC protocol on VolTRAX for four samples with reagents pre-filled in the cartridge.

VolTRAX should have an API released this summer to those holding developer licenses, allowing custom protocols.  A new user interface is being rolled out as well.

Focused Applications

LamPORE will ultimately be a kit, targeted for August.  ONT did not reveal much of their regulatory strategy, but presumably this will be coupled with appropriate filings in the U.S., U.K., E.U. and elsewhere.

ONT is promising their Cas9 enrichment will be available this month.

Adaptive Sampling

Perhaps one of the most exciting announcements is that MinKNOW will soon support adaptive sampling, formerly known as read-until, on MinION and GridION. By supplying a reference sequence and intervals on it, one can either enrich or deplete specific regions of a target by actively ejecting molecules that don't fit the user-defined criteria.  This capability has long been in the platform, but required developers licenses.  Now it is going mainstream -- but not yet on PromethION, though that is in development.


ONT has been trying a number of approaches to improve the accuracy of the platform. As noted above, the higher accuracy R10.3 pores will soon be available in all three flowcell formats.

The school of research basecallers has not been culled.  Guppy, Bonito and Run Length Encoding are all out there.  A new "rerio" site will make available more basecalling models in Guppy format. 

Bonito now offers a pair decoding mode which can take two files known to represent clones of the same molecule -- associated by UMIs perhaps -- and generate a higher quality  consensus that using a tool like Medaka after aligning two sequences.  

Medaka still doesn't have a mode for combining R9.4.1 and R10.3 data.  Nor was there talk of alternative pores, using base analogs or the return of 2D chemistry.  Impossible to tell if these missed a cut for time considerations or they really have been back-burnered or axed.

Pundit Time

Given the constraints of the pandemic, ONT continues to move the platform forward.  But, there are a lot of directions going on and it isn't clear how many of these will ultimately come back together.

For example, on the flowcell front we already have three flowcell form factors, two pore types and two library preparation kits.  In the past, these have never formed a full combinatorial set of options -- this kit won't work with these pores or these pores aren't available in this flowcell form factor.  There are further interactions: VolTRAX library kits seem to all be around transposition chemistry, which often seems to be late to the party when it comes to new flowcells

Many of the new developments risk further fragmenting the platform.  The low idling motor options appears to be available only for ligation.  Which flowcell formats and pores will get the extended mediator treatment?  There's also the question whether anything potentially altering the local pore environment, such as the modified wells for more mediator or changing the mediator mix, might require new basecalling models.  Can the extended runtime designs be combined with the high sensitivity designs, or are these going to butt heads with each other?  Which library prep methods and flowcell formats will be supported on the high sensitivity devices?  

Plus down the road is coming the denser pitch devices.  Will these be entirely used for new form factors such as SmidgION and Plongle or will they also be made available in the older formats?  Of course as a user I'd like to see all the formats, and perhaps even more. I dream of PipettION, which is just a fancy tip for an Eppendorf, or a sticker-like SmudgION that could easily be built into other hardware.  Managing manufacturing and distribution complexity has never been Oxford Nanopore's forte though; higher complexity without premier execution is not a recipe for success.

No comments: