Thread by @ewanbirney

For new followers (and with apologies to long standing followers), I am a long established paid consultant to Oxford Nanopore - a DNA and RNA sequencing company with a very different sequencing chemistry from previous chemistries. Some background:

I've consulted for ONT for over 10 years now (!) and seen the twists and turns of both technology development and commercialisation in a complex environment. There have been some serious twists and complications.

Like a lot of successful technologies, many people need to be credited with its success - original academics (David Deamer, Dan Branton, George Church, Hagen Bayley ...), the business people who saw the opportunity (Gordon Sanghera, Spike Willcoxs)+ scientists inside ONT >>

I know the scientists inside ONT the best - Clive Brown (@The__Taybor), James Clarke, Rosemary Dokos, Stuart Reid, Tim Massingham, Chris Seymour and others and there is huge amount of simply awesome science *inside* the development/company process.

A good example is that much of the smarts of the device is not about the nanopore, it is about the lipid bilayer and the device which holds it. As Clive wryly points out "we're shipping soap bubbles around the world and they have to work at the other end"

Then there is massive drama about the motor. Nanopores in the 1990s was famously "the sequencing technology for the future, and always will be" because of trying to control the movement of nucleic acid through the pore.

There is a whole sideshow of silicon meets biological materials and just how fiddly a device which can measure current through a single molecule is to all sorts of little differences (buffers, robustness, QC...)

Finally decoding the signal is ... quite a drama (as Chris or Tim will point out). It is not a simple signal, and although explicit, generative models (HMMs and friends) are good, Deep Neural Networks have been key in leveraging the signal

(not least because one has both amplitude and time-domain signal, convoluted across multiple base pairs and in some case spanning the motor position some 16 base pairs away).

...but... then... you get something that sequence DNA or RNA, with the length of read only limited by sample prep (yup, you read that right, there is no *inherent* limitation to the length of nucleic acid polymer through a nanopore) on a set up which is the size of a stapler.

It can read DNA or RNA; it senses DNA and RNA modifications "as if they are a new base" (cue - huge training and informatics headaches, but a nice headache to have - suddenly the RNA modification world is accessible to us)

It is fast (first reads off in 30mins), asynchronous and very portable. People have sequenced in space (ISS) (!!!), in Antartica, in Costa Rica jungles; sequencing in GPs practice should be ok :)

What's not to like? Well... skeptics will point at error rates, which have been steadily dropping with better pores (R10.3 better than R9 series), better movement control in buffering etc and better informatics; yet more improvements are coming down the pipe

Although there is just one molecule of DNA going through the pore there are many ions traversing in each time-slice of sensing so there is not a stochastic sensing problem; there is a single molecule modification/state/conformation issue which is a blessing+curse

Quite where the final single molecule limit is I don't know - but it has a way to go. But then as soon as one gets into consensus building in a variety of ways (double strands, UMIs, assembly haplotypes...) the situation is really sorted. Plenty plenty more headroom.

I've really enjoyed hanging out and helping out where I can the nanopore crew (Doffs hat), and it has also inspired some new science from my group / associated with me.

My two current favourites are RNA modification detection (With Tommaso Leonardi and Adrien Leger, biorxiv.org/content/10.110…) and Adaptive sampling with Nick Goldman and Matt Loose groups (biorxiv.org/content/10.110…)