Sometimes this genome blogging lark really pays off. Yesterday was one of those days as I got a sneak preview of the big announcement at AGBT, and 20 minutes to speak with Oxford Nanopore’s Dan Turner (Director of Applications), Clive Brown (Chief Technical Officer) and Zoe McDougall (Director of Comms). The downside of course was that I couldn’t tell anyone what they said until the embargo was lifted at 5pm today!
I asked as many questions as I could without knowing the contents of the AGBT talk. I probably should have asked a bunch more. I do remember saying “wow” quite a lot.
First, go and read the press release!
- Nanopore have announced a strand sequencing method, made possible by a heavily modified biological nanopore and an industrially-fabricated polymer
- DNA passes through the nanopore and tri-nucleotides in contact with the pore are detected through electrochemistry
- Demonstrated 2x50kb sense & anti-sense of same molecules (lambda phage) – no theoretical read length limit
- Can sequence direct from blood without need for sample preparation
- Two products announced:
- MinIon – USB disposable sequencer for ~ $900 has 512 nanopores – target 150mb/hour
- MinIon can run at 120-1000 bases/minute per pore for up to 6 hours
- GridIon – two versions of rack-mountable sequencer with 2000 nanopores (2nd half 2012), 8000 nanopores (2013)
- GridIons can be racked in parallel, 20 could do a whole human genome in 15 minutes
- Each GridIon can do “tens of gigabases” over 24 hours
- Both machines commercially available 2nd half 2012
- Sequencing can be paused, sample recovered, replaced, started again
- Accuracy is 96%, errors are deletions, error profile will improve through software
Firstly you need to know this is pronounced “min, ion” rather than “min-yon”. The MinIon is an array of nanopores – 512 to be precise – and circuitry housed in a USB stick. Why a USB stick? “The form factor is determined by the requirements” – as there are no fluidics you don’t need a big machine. There are no fluidics. “Your fluidics is a Gilson”, said Brown. The prototype version has an ugly battery pack attached to it but it will eventually use USB power. The USB stick is disposable. “Why do you need an instrument?” he says. We wander into the realms of sci-fi at this point. DNA molecules pass through the nanopore and nucleotide sequence is detected by the electronics. Bases are streamed – live – to your laptop as FASTQ (bases with qualities). This is where the “run until” makes sense, if you are interested in a particular gene just wait until the sequence comes out and shut it down to preserve the circuitry.
“Hungry Hippo” sequencing
The sample lives in solution and the DNA floats around, there is no attachment to beads or solid surfaces. Brown says the nanopore functions “like a Hungry Hippo”, grabbing DNA as it bumps up against the pore. Only about 0.001% of the starting DNA is processed and the sample is unadulterated and can be recovered at the end. Do some sequencing, then literally “pause” it, remove the sample – mess around with it – replace it and continue sequencing. This is an iterative approach – how will it change our way of doing sequencing experiments? Can you run forever? “Right now this is limited by the circuits which burn out after about 6 hours.”
Nanopore can ratchet down the speed of the DNA passing through the pore. Right now they are running “about 120-150 bases per minute”, but this can be sped up to give 150 megabases per hour of sequence from the MinION. It can be sped further, even as far as 1000 bases per minute per pore, although running above 500 bases per minute starts to impact the accuracy.
Read length reflects the fragment size in the sample. They’ve sequenced sense and antisense of lambda phage – giving a 100kb read from same molecule. “We’ve tried 5, 10, 15, 20kb etc. without problems” says Dan Turner. You could bung an Illumina library in there should you wish.
A huge question. They are talking about a 4% raw read error rate. The errors are mainly deletions. “We know what the problem is and we can fix it” said Turner. He explained the signal comes from the interaction of DNA with the pore barrel. The barrel is in contact with three nucleotide bases at one time, then the strand is moved along one base by a “special processive enzyme”. Each tri-nucleotide gives a specific signal. A regular Viterbi algorithm turns that into bases, and the Markov model can be improved through better training. Notably the quality does not drop off over the length of the read and substitution errors are infrequent.
A major problem with turning biological nanopores into DNA sequencers is that they are often embedded in fragile lipid bilayers which are unstable to pH and temperature (see James Hadfield’s great nanopore primer for more information). The major break-through is that Oxford Nanopore can make this membrane in an industrial process in a factory and the result remains stable, meaning it can be shipped to the customer at room temperature. The nanopore itself has over 300 modifications to make it work efficiently, these were found through random mutation during a series of high-throughput screening experiments.
With the MinIon – bases are streamed off in FASTQ straight to your laptop. You can set up your workflow to “run until …” a particular sequence is read, and then shut it off to protect the circuitry.
The GridIon has already been announced and is a rackable sequencer which takes disposable cartridges for massively parallel sequencing. It will be available in 2000 and 8000 nanopore versions. There are four wells to one circuit.
The example given at the presentation was sequencing of whole blood. “You add blood, and some buffer and some enzymes”, presumably to nick the DNA strands which don’t need to be denatured. That’s it. But you could process your sample in a million different ways first, as long as dsDNA is presented to the pore. “You could even put an Illumina library in”. No amplification required. At the presentation Brown said they “had a look at RNA” and found they could read it just like DNA, no reverse complementation necessary.
The Illumina Connection
This is an Oxford Nanopore product, being marketed by Oxford Nanopore. Illumina have a stake in the company.
Is this a game changer?
This is a very different product to the existing suite of sequencers. It’s also rather dissimilar to PacBio despite being single molecule (but apparently superior in all respects).
For those chasing the goal of doing tens of thousands of $1,000 human genomes, this may not seem to be the obvious solution. The cost per base is competitive with existing systems. But I think we need to think differently about this machine:
- how will access to a disposable sequencer change the way we do biology and medicine? With no capital costs, this certainly has the power to democratise sequencing even further. But does the $900 price point make it a tough sell for near-patient testing or field applications, or will clever use of the technology make it economical? Immediately you imagine that you could load a sample, sequence for 10 minutes, remove, put another sample in to get a kind of barcode-free multiplexing
- how will 100kb+ reads change your research? It seems to me this will solve the very gritty problem of trying to reconstruct microbial communities through assembly of short-reads. Large chromosomal variations should be detectable with ease. Certainly de novo assembly will become trivially easy for many organisms. Should be amazing for ‘unculturable’ microbes too.
- how will “run until” change your research? If you are looking for something specific, you just wait until you have got that particular bit of chromosome 7 running through the pore and stop – no need to oversample the whole genome to 100x.
- we can now take the sequencer out into the field. How will this transform microbial ecology? You can go and pipette in some sewage water and be a 21st-century John Snow and detect cholera. Does this help with the “disease weather map” ? You can imagine a sexy Kate Winslet / Marion Cotillard style epidemiologist armed with one of these and a laptop.
- what experiments can you do with iterative sequencing? What are the benefits of taking a sample on and off and manipulating it? Will this change the way we do RNA & ChIP-Seq type research?
I’m still processing this announcement and would love to hear your thoughts below.
But to sum up, this is the megaton sequencing announcement we’ve been waiting for.