Ewan Birney Profile picture
May 21 34 tweets 7 min read Twitter logo Read on Twitter
Here is the slightly cheesy montage for the great #nanoporeconf for 2023 - and, with a reminder of my conflict of interest - I am a longestablished paid consultant for Oxford Nanopore and a shareholder - here are my thoughts on the conference.
For long time nanopore scientists -and I am definitely one of those- one can definitely both plot progress London Calling conference (on the Thames in London) both in terms of what the company presents as near and long horizon+how the plenary speakers use and talk about nanopore
From the company side, much of this was giving a roadmap of key software and flow cells; the R10 flow cells (which is a distinct step up in quality) are now routine; what is not yet is high yield duplex which has being moving from Oxford to Alpha to broader Beta testers.
As well as the high yield duplex flow cells (some magic in keep the second strand in the right place to follow the first) as important is making the duplex calling one-button push; currently one needs a rather complex workflow to find and recall the duplex.
The high yield duplex revives an old (2018? 2019?) idea to have a synthetic second strand protocol with spiked in modified bases of the company's own design.
This should solve the homopolymer issue as the modified bases from a DNA polymerase perspective are incorporated against a homopolymer randomly, but they will give changes in nanopore signal. With R10's long read head one only needs 1 in 9 - say 1 in 5 in practice to get length
Talking to some of the "nanopore wizards" (JC and Stuart) afterwards they are bullish about this, as is obviously Clive (otherwise he wouldn't have shown so much). Clearly it is fiddly (what bit of single molecule sequencing is not) but clearly do able.
This means, with exisiting nanopore tech (no need for outie chemistry for example) one can see a path to Q20+ single read, Q30+ duplex *including* homopolymers, and frankly it should start getting hard to assess consensus quality as it goes through Q40 onwards
(past Q40 - 1 in 10,000 error rate, goint to Q50 - 1 in 100,000 it gets ever so hard to have big enought truth sets with enough accuracy and consistency- you have to know the right answer crudely at least 10 fold better - so 1 in 100,000 / million errors on your truth set)
Another big thing are the new chips (ASICs) which I can see the nanopore team are excited about. It is the first ground up redesign of the chips, and uses less silicon per channel (no more of these 4 muxes - they have gotten better at putting in one pore per membrane)
There are many other tweaks, most of which I could not follow but sensed the collective excitment. The end results is more channels per silicon, far, far less energy (really bring in mobile applications) and integrated (so cheaper to make all in)
I'm not sure quite when we will see products based on these, but they already committed that some of them will be compatible with MinIONs (so these will be "just" a new type of flowcell). Low power rating means true mobile, hands off applications are conceptually doable
(one still have to sort out the "front end" - molecular biology - or "molly-bolly" in internal slang. One option is the revamped VolTrax - now named something else. I wonder if for blood and water/or simple inputs there could be just a bead/column/simple attachment thing
This was a 2019 idea - ubiquity or something - and clearly they can get some of this working. I think there will be multiple ways to get the front end to work, ideally "technician hands free", but this is an active area of research)
The final theme for me was modifications, both DNA and RNA. As long known and expected (I remember I think back in 2015 or 2016 saying that CpGs can be read directly) nanopore can sense many modifications; what has changed is a robust training mechanism
So now, already, R10 nanopore called native DNA can have CpG methylation and CpG hydroxymethylation (it is going to drive us all up the wall, but the most common way to get methylation read out - bisulphite conversion - actually converts both of these. Sort of union CpG methyl)
Here the hidden world of RNA I think will open up - RNA is not some one shot, passive molecule either structurally doing things (rRNA) or passing information (mRNA); rather it is an actively managed molecule in cells, being marshalled, sorted, translocated and degraded.
We've known for a long time mRNA is super dynamic; we've known for a long time mRNA is modified; we've not had the technology to look at how modifications either are deliberate marks of their management or "just" reliably bystanders on their journey.
On a personal note, it is great to see one of my ex-postdocs, @AdrienLeger2, leading this area in Oxford Nanopore and a good example of academia to industry transition and fun.
Turning away from the company to the presentations, unsurprisingly it was more on the production schemes available over the last year, ie, R9 and R10 simplex (with some R10 duplex), but the innovation of how to use nanopore was going up
Some of this is obvious - long assemblies (are long reads - and really long reads - useful for assemblies - of course they are!). Here the "nanopore only" assemblies closing in on the "best blend" and being good enough for many applications was interesting
(I thought the NIH Alzheimer talk by Kimberley Billingsley of 200 human genomes from brains, being able to get good diploid assemblies with methylation and some somatic mutations was an excellent talk).
The other obvious use of nanopore is in cancer, where the combination of long reads for structural variation, methylation+hydroxymethylation (and more in the future) "for free" and fast turn around is going to be great
I am biased as I know the work from @hbelrick on this from @emblebi and Helen Webb's talk from @GenomicsEngland (also an area I know) was good to see getting out of R&D and closer to production
(someone moaned at me that it wasn't yet there for GeL, which is... naive about the difference between "works for clinical research" - we've seen plenty of this - to "is a routine offering in a healthcare system". They are related things - but not the same!)
One surprise to me was the utility of long read single cell transcriptomics with multiple talks on this. I had stupidly thought that the only real benefit of long read transcriptomics was splicing - and the bulk of the information was in the expression.
However, both Rachel Thijssen and Hanlee Ji made huge use of linking the mutation (somatic cancer, or engineered) to the transcript-cell level bar code to immediately be able to track specific mutation to cell barcode. this is ... big.
I can see some really interesting use cases of this, and it removes the need for one layer of index tracking potentially in engineering (when one is dealing with transcript mutations, but those are some of the most important ones).
In general the long reads being able to link information - whether it is SNPs to Methylation, or mutations to bar codes, and I think in the future things like artificial modifications (think... chip adaptations) to SNPs, CRISPR sites or between types it is going to be... great
Finally adaptive sampling - controlling how the chip accepts or rejects nanopore reads- is coming good. Close to my heart is @weilgunyl's BOSS-RUNs (dynamic adaptive sampling) and in the future dynamic adaptive assembly.
But adaptive sampling is both empowering for routine uses (eg, "just the exome") through to this more sophisticated lets update the sampling strategy as we see the data, which I think will be useful in broader and broader applications.
So, as ever the future is bright for @nanopore (when has it not been - steady progress each year) with the endless utility of measuring DNA (and RNA) - both natural and manipulated/engineered in a variety of ways.
There is plenty of headroom in the key features - such as error rate (nanopore is nowhere near its maximal), speed (silicon density really) and sensing (eg, modifications, already realised, "just" needs training) and plenty of future in its use.
A final reminder - I have a conflict of interest as a long established nanopore consultant and a shareholder. No company vets my tweets.

• • •

Missing some Tweet in this thread? You can try to force a refresh
 

Keep Current with Ewan Birney

Ewan Birney Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @ewanbirney

May 21
My friend and economics/ markets guru @felixmwmartin commenting on super human AI and all too human market behaviour - on the money that AI will transform many things (science included - it has started in earnest) but also more broadly in the economy
Economics and biology are closer in data science than you might think - in particular micro economics and observational human biology aka epidemiology. Plenty of differences but lots of overlap as well, eg biased sampling, many hidden confounders, clearly correlated variables
A deeper issue is the need to understand causality / intervention- if I enacted this policy or provided this drug what would happen next. Finding the golden causal threads in the tangled Gordian knot (hairball?) of correlation is a common challenge shared by biology+economics
Read 6 tweets
May 11
The @HumanPangenome - the 3rd chapter. For an intro into the human pangenome, check out this thread - - and additional papers - but now on to the weirdest "our genome does that!" moment. Brace yourself for genome geekery
We, like most other organisms, have a particular thing when we make sperm or eggs (depending on our sex) - we have an enforced "shuffle the deck" our parent's genome - this is called crossing over or recombination in meiosis.
In normal cell division, our 46 chromosomes duplicate, make little pairs (the fat X's in many graphics), line up at the equator of a cell, and each pair is pulled apart to each pole. This scheme ensures that nearly always each daughter cell gets a precise copy of the genome
Read 16 tweets
May 10
It was a tour de force just generating a draft human pan-genome @HumanPangenome, but some suspected and some utterly bizarre things about the human genome have been revealed using these 47 different human reference quality genomes.
For a primer on human pan-genomes, here is a thread that will take you through this. Read this first before going into the weird and wonderful world of our genome.
Some things the researchers expected. One was how to just routinely handle this level of the data. For a long time computational biologists have known that "something graphy" was the right thing to keep track of many different things. The practical end of this is... mind bending.
Read 15 tweets
May 10
Congratulations to the @HumanPangenome - in particular @BenedictPaten for his leadership + highlighting @ensembl's role via Fergal Martin in annotation. For the rest of this thread I will explain this technical genomics tour de force using... Shakespeare nature.com/articles/s4158…
Shakespeare wrote a considerable number of pieces of work; English was a very fluid language and so his own spelling was quite variable and then subsequent copying errors and improvements have led to many differences in the documents representing this work.
When Shakespearean scholars want to comment on a line of a play they need to agree on a "reference edition" - say "the 5th Arden Edition" - just so that if there is quibble about the precise phrasing - "spotless" or "spotles" scholars know they are talking about the same thing
Read 19 tweets
Apr 16
This is part right (“we’re on a cusp of further genetic understanding of health etc) and then straightforwardly *wrong* on its implication (“there are inherent differences between groups”) and this pseudoscience twoodle by @GoodwinMJ needs … debunking >>
Basically humans don’t come groups - we exploded across and out of Africa very quickly (in evolutionary time) and we’ve moved and mixed ever since - in antiquity as well as now
This means human genetics is a big complex family tree (formally a set of trees across the genome) and ultimately is just very messy. Unlike other species who adapt to environments by genetics we mainly have not, rather we’ve done far far quicker via behaviour
Read 10 tweets
Apr 15
A saturday morning muse triggered by a great random meeting of the Norwegian cohort community in Bristol, due as ever to the pro-social @mendel_random and @carolinerelton, on the convergence of population health genetics to clinical genomics.
Some history - keeping track and analysing cohorts of humans has a long pedigree in science, and is one of the fundamental tools in epidemiology. There are two things one has to addree (a) logistics in recruiting and tracking dominate cost and effectiveness (b) ascertainment
By ascertainment one really means is the set of humans you are tracking a random draw (usually of the population) such that the inferences you make from the population you can generalise to everyone.
Read 23 tweets

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

:(