Oxford Strikes First in DNA Sequencing Nanopore Wars

February 17, 2012

By Kevin Davies  

February 17, 2012 | Breaking a near total vow of silence after three years in stealth mode, Oxford Nanopore has offered stunning details of its nanopore next-generation sequencing (NGS) technology to a packed house at the premier genome sequencing conference in Marco Island, Florida.  

In a talk at the Advances in Genome Biology and Technology conference this morning, chief technology officer Clive Brown presented details of what the British company calls its “new generation” of single-strand sequencing technology, featuring accurate long reads of single-stranded DNA molecules, including the completion of two viral genomes.  

In the second half of this year, the company will commercially release its previously announced GridION instrument, or node, as well as the world’s smallest DNA sequencing instrument – a portable cartridge called the MinION that doubles (and works as) a USB drive.   

 Oxford Nanopore’s disruptive MinION USB device   

And in a presentation that only briefly indulged in marketing hype, Brown laid out the prospect of achieving the $1,000 genome in 2013 in under an hour by harnessing 20 GridION nodes working with a second-generation nanopore array. 



And in a presentation that only briefly indulged in marketing hype, Brown laid out the prospect of achieving the $1,000 genome in 2013 in under an hour by harnessing 20 GridION nodes working with a second-generation nanopore array.   

Brown and Oxford Nanopore CEO Gordon Sanghera gave Bio-IT World a preview of today’s announcement earlier this week. Among the system’s notable features are:  

  • Long uninterrupted read lengths of uniform accuracy from beginning to end  
  • Arrays featuring 2,000 bespoke nanopores in a proprietary polymer bilayer coupled with a custom processive enzyme that ratchets the DNA template through the pore one base at a time  
  • An informatics system that reads overlapping nucleotide triplets in the middle of the pore to deduce the sequence  
  • The prospect of a sub-60-minute human genome in 2013 by networking 20 GridION nodes in unison.  
  • By sequencing long single strands with high (if not yet perfect) accuracy, competitive cost and minimal sample prep, Oxford has taken a giant leap not only in the nanopore wars but also the entire NGS field.  

     “When we sat down to design the system four years ago, I said we have to compete on every factor,” says Brown. “If we get half of them [sorted], we’ll win. And we’ve actually got almost all of them… Too many companies try to compete on one differentiator and they often fall short. We’ve tried to compete on every differentiator – a deliberate choice on our part.”  

    But Brown and colleagues are not basking in celebration just yet. Asked if he was impressed or satisfied with the progress his team has made over the past 1-2 years, Brown stated: “There’s no jubilation from me – this is just a starting point, not the end point. It’s the next year that matters,” referring to the early access program about to get under way, followed by commercial launch later this year.  

    [Editor's Note: Clive Brown will present the latest on Oxford Nanopore's sequencing platform at the inaugural Bio-IT World Asia conference in Singapore this June.] 

    The Basics  

    Clive Brown, Oxford Nanopore CTOSince the mid 1990s, nanopores – natural bacterial proteins that punch holes in cell membranes – have shown immense potential for DNA sequencing and a plethora of other biological sensing applications. Oxford Nanopore was founded by Oxford University chemist professor Hagan Bayley, an authority in nanopore biochemistry.   

    In 2009, Oxford Nanopore researchers published a landmark paper describing the ability of a natural nanopore to discriminate the four nucleotides in DNA by measuring electrical current blockage, providing the basis for a sequencing strategy involving an exonuclease to cleave individual bases of a DNA template that would fall into the pore to be detected. Illumina invested in the UK company, becoming a minority shareholder.  

    But in the past 12-18 months, advances in single-strand DNA sequencing, some coming from academic collaborators such as Mark Akeson, professor and chair of biomolecular engineering at the University of California, Santa Cruz, offered a more rapid route to commercialization. A key factor was to complex an enzyme with the nanopore to harness and control the speed of the DNA strand as it traverses the pore. “Akeson provided the proof-of-concept, we really picked up on that,” says Brown. “Once we’d replicated that, everything after that has been completely Oxford Nanopore.”  

    A year ago, Oxford Nanopore released details of its GridION instrument, or node, but said no more publicly about its research progress. Brown espoused the same fixed focus ten years ago while at Solexa, the British NGS company acquired by Illumina in 2007.   

    Over the past 18 months, Oxford scientists led by John Milton studied hundreds of nanopores for optimal properties from about a dozen porin families, including some 300 in detail. “We have a great pipeline for cloning, expressing and mutating nanopores,” says Brown.  

    The company settled on a custom engineered, rather than a wild-type protein. “We’re not using the academic system,” says Brown. “People tend to assume that’s what we’re doing, we’re not. We found our own nanopore and our own enzymes.”  

    Indeed, Oxford is also using a special enzyme to process the DNA, not the DNA polymerases currently used by some academic teams. (Brown would not disclose the identity of the enzyme.) The enzyme is capable of ratcheting DNA through the pore at upwards of 1,000 bases/second, a rate that is controlled by various cofactors in solution. The company is currently settling around 20-400 bases/second.  

    The bilayer membrane is also a synthetic proprietary polymer, rather than a natural lipid bilayer. This gives Oxford “incredibly robust bilayers,” says Brown, allowing the company to use biological samples (for example raw blood) that would normally destroy a natural bilayer. “So we can look at very dirty samples,” he says, including sewage-tainted river water running near the company’s lab on the outskirts of Oxford.  

    The first array chip, available later this year, contains 2,000 nanopores (2k). An 8,000-nanopore array (8k) is planned for release in 2013.  

    Going Viral  

    Rather than laying DNA on a solid substrate, the Oxford Nanopore system works in solution. Each of the arrayed nanopores independently grabs a DNA strand in real time. “The pore sucks [the DNA] through like spaghetti, with no degradation of signal,” says Brown. “The signal at the first base is same as the last base.”  

    Brown presented results of two viral sequencing efforts. The first was “an old friend, not a particularly glamorous genome” – PhiX174 – the first genome assembled by Brown and colleagues at Solexa a decade ago (and the first organism ever sequenced, by Fred Sanger in the 1970s).   

    Unlike his time at Solexa, when Brown issued an email to senior management proclaiming “We’ve done it!” when the first virus was sequenced, Oxford’s achievement was “very incremental.” “We’ve had long reads for a while,” he says. “There wasn’t really a eureka moment as such that I can pinpoint.”  

    The Oxford Nanopore system comfortably handles the entire genome in a single read. By engineering a hairpin at the end of the molecule, both strands of the template DNA can be sequenced in succession on the same pore.   

    From there, Brown’s team went onto sequence the genome of bacteriophage lambda, about ten times larger, around 48 kilobases. Again, the pores could sequence the entire linearized phage genome in a single run. “Up to 100-kilobase reads have been accomplished on the platform, that’s not exceptional,” says Brown.  

    Brown puts the raw error read rate at 4%, which he views as satisfactory for the time being, but because of the throughput and large read lengths, there is a “massive consensus” on the final sequence. Future improvements in accuracy are likely to emerge from improvements in engineering and sample prep methods.  

    “We know what causes errors – a defect inside the pore causes a bit of noise,” says Brown candidly. “Going forward, we’ll just engineer the pore to remove that noise. An incremental improvement in pore development should take us to a lower noise system.”  

    Most of the errors tend to be deletions, which won’t significantly affect alignment or assembly, because they will leave a gap where the bases were. Sequencing homopolymers – a major issue in some NGS platforms – also doesn’t appear to be a major issue, in part because Brown says the enzyme that processes the DNA has what he terms a “movement tick,” providing a built-in counter of sorts.   

    A key feature of the system is that there is no degradation in read quality from beginning to end of the fragment, unlike most other commercial systems, where there is a decay in quality scores towards the end of each individual read length.  

    Brown also touched on additional features including RNA sequencing and detection of epigenetic modifications in DNA, but details on those aspects will follow later.  

    Informatics Issues  

    At the heart of the informatics processing in the Oxford system is an ASIC (application-specific integrated circuit). A vital component, says Brown, was to design “a sensor chip that does electronic readout” in order to produce parallel sensing on a sufficient scale and high quality. “You’ve got to make a custom chip. They must be designed from scratch.”  

    This serves as a low-noise amplifier, a digital-to-analogue convertor on a parallel chip that sits underneath the biological chip -- a series of wells covered with the polymer bilayer, into which the pores are assembled.   

    The actual bioinformatics around decoding the sequence is not, as many had anticipated, detecting individual bases in the center of the pore. “What everyone has been striving for is single-base resolution,” says Brown. “We figured out you don’t need to do that.”  

    Brown’s group has opted to read successive Kmers in which K is a triplet. (The signal block could range from 3 to 11 bases, depending on the pore.) As the enzyme unidirectionally moves the DNA strand one base at a time, the current corresponding to each overlapping Kmer, or 3mer, is read.   

    The signal processing technique involves a Hidden Markov model and a Viterbi parse finder. It’s a classic informatics strategy used in chess programs, says Brown. “We calculate the [possible] paths and condense out the base calls.”   

    While Brown doesn’t rule out a single-base reading system, he says, “There’s a fair chance that single-base resolution reading (a) might be a long way off and (b) might never be better than [the current system].” On the other hand, reading longer Kmers might be better, but may not be a necessity.  

    The sequence is produced as a streaming fastQ file. “All software issues are dealt with on the instrument, and we do the bioinformatics -- alignment and assembly -- in real time, instead of waiting for the end of run.”  

    Oxford has a previously announced collaboration with Accelrys and helped develop an NGS application for its popular Pipeline Pilot software.   

    Sequencing on a Stick  

    Perhaps the biggest surprise in today’s announcement is the creation of a sequencing device so small it actually doubles as a USB stick. It will retail for less than $900, and should pro competitive with Illumina’s MiSeq and Ion Torrent’s PGM instrument.   

    “The minION breakthrough means we can premake the bilayer in the factory, put the pores in. Then you’ve got a shippable system,” says Brown.   

    The memory-stick sized sequencer contains one ASIC and 512 channels, capable of producing about 120 megabases sequence per hour. The device literally plugs into a USB port on a laptop or desktop.   

    “You just pipette the sample in, using a Gilson pipette -- your thumb is the pump,” says Brown. Once the sample has been loaded, sequencing beings. An app on the computer drives the device, once again producing streaming fastQ files in real time. The reads are picked up by Pipeline Pilot in a premade workflow.    

    “You can look for things in real time,” says Brown. If the power is shut off, the sequencing reactions stop. An experiment can be suspended, the sample recovered, treated in some manner, and sequencing restarted.  

    “The whole workflow is completely accessible to any scientist,” says Brown. “You can put your arms around the entire problem, without having to hand off to library prep groups and cluster growers and machine runners and bioinformaticians. You can put your arms around the whole problem.”  

    Because the minION is small and sealable, it is ideal to be taken out on the road. All that is required is input of DNA with an overhang. Brown sees a host of potential applications from developing countries to the battlefield, environmental monitoring to educational teaching.  

    Price & Performance  

    Oxford is not announcing a price for the GridION node itself, but will offer five different price bundles, in which the cost/Gb “will be competitive to current incumbent offerings,” according to CEO Gordon Sanghera. The goal is to provide flexibility to meet customer needs, depending on whether for example they prefer to minimize capital expenditures and pay for service, or a willing to pay higher prices for sequencing without so much upfront capital investment.   

    “We want to leverage the way you price a mobile phone,” says Sanghera. Some customers prefer a service plan, others prefer lower upfront costs and to pay as they go. “We’ll show transparent pricing on all our products,” says Sanghera. “All customers will be treated equally.”  

    The price guidance for the 2k array is $25-40/Gb; the 8k array (in 2013) around $25-30/Gb. The MinION will cost $500-900/single use. “The MinION is a fantastic extrapolation of scalability,” says Sanghera.   

    Oxford Nanopore is banking on requiring only a modest engineering field force or sales force. Most of the ordering and fulfillment will he handled online.   

    In terms of performance, Oxford says that the cost/base for the GridION will be competitive to Illumina’s HiSeq, and has the potential to reach the $1,000 genome in under an hour by 2013.   

    Throughput scales with the number of pores X the speed of sequencing X the time of the run. A network of 20 GridION nodes running the 8k arrays at 300 bases/second could in principle deliver 43 Gigabases sequence in 15 minutes, or a $1,000 genome in under an hour. (The company says the sequence will cost less than $10/Gb.)   

    How many GridIONs can be networked together? Brown says there will be a practical limit,  but “it’s a fairly big number. All the data reduction is done on board. Because [the dataare ] streamed, it’s very non-lumpy I/O. Your cluster disk storage will like it a lot, because there’s no ‘lumpiness’ in the I/O sense.”  

    With a path to high-throughput, single-molecule sequencing with ultra-long reads, good accuracy, low costs, and easy sample prep, Oxford’s announcement raises the stakes for many other nanopore sequencing companies, including NABsys, Genia Technologies, and NobleGen – if not the entire NGS field.   

    “Some of them are blatantly copying us, and some have wacky ideas that I don’t understand,” says Brown. “As Gordon [Sanghera] says, the nanopore war is about to start. There’ll be a lot of shouting and grandstanding in the next few months. But in my view, it’s like Solexa: we just have to get on with it. Credibility comes from experience in the field. That’s what we have to crack next.”