Oxford Nanopore’s Clive Brown Looks Back and Ahead

June 4, 2024

By Allison Proffitt 

June 4, 2024 | At Oxford Nanopore’s user conference—London Calling—Clive Brown got a bit nostalgic, recalling his 2012 presentation at AGBT, “outing” Oxford Nanopore’s technology, product plans, and use cases. “I knew the talk would get attention, but I didn’t think it would get that much attention,” Brown remembered. “There was a whole furor afterwards about the talk.”  

The Oxford Nanopore vision was of a “new generation of single-strand sequencing technology, featuring accurate long reads of single-stranded DNA molecules,” Bio-IT World reported from the 2012 event.  Today, Brown’s big ideas include whole-chromosome sequencing, flow cell loading without a pipette, and advances in chemistry, electrical engineering, and commercial format.  

From the launch of the MinION, Brown said, Oxford Nanopore was committed to disruption. “Once you go down that very disruptive route, everything else has to be disruptive,” he said, including technologies, marketing, ways of connecting with customers, and pricing schemes. That approach led to the launch of the London Calling user group meeting. “The features should sell the product,” Brown said, “And the imagination of the customer should be the driving force.”  

Those early days of the company were also “really difficult,” Brown said. “I think we had three lawsuits at the time, at least three large companies trying to crush us. As I gave the AGBT talk in 2012 I could see people scribbling notes and then they filed concept patents that they then used to sue us.” He continued: “That period was very, very difficult. Still have PTSD from it.”  

Oxford Nanopore Today 

Today, Brown does not seem to be suffering too much from past trauma. Nanopore is “all you need,” he told the users attending the event. The latest versions of the technology include the newest pore (R10.4.1) and helicase motor (E8.2.1) running at 400bps (bases per second). Simplex sequencing is available in store; duplex sequencing for higher single-molecule accuracy launched about a year ago and is available in early access.  

The platform today, at 30x coverage, is routinely seeing 99.6% modification calling accuracy, 96.1% structural variant calling accuracy, and 99.7% single-nucleotide variant accuracy, Brown reported. ONT has worked with Clair3 to improve short variant calling, improving SNP accuracy to 99.9%  

Researchers are using Oxford Nanopore devices all over the world—and beyond. From deserts to jungles to the International Space Station and submarines, Oxford Nanopore MinIONs are in the hands of scientists. “I would emphasize that these devices are still not lab free. So growth for us comes from getting more scientists to these sequencers who wouldn’t normally do it. It also means getting sequencing out of the lab and into applications that don’t require laboratory workflows. We’ve not done that yet; that’s a work in progress.”  

Nanopore is pushing platform performance, Brown said, in every area, from sequencing chemistry to sample prep to motor design to platform electronics to analysis. “A beauty of nanopores is you can fiddle with run conditions. You can fiddle with speed, salt. It’s a very plastic system, and that means you can get complicated but very elastic signals out. And that means you can walk up an optimization ladder.”  

For example, the motor Nanopore uses to walk bases through the pore has improved dramatically in speed, starting at 30bps now up to a default run speed of 400bps. But that isn’t an upper limit, Brown said. He was challenged by a teammate to go faster. “It was a very debatable point at the time because of the time-domain behavior of the motor,” Brown said. “What happens if you run quicker? So we did that, and the system did not fall off a cliff and maintained a reasonable degree of accuracy. We’ve walked up that ladder ever since.” 

Basecalling and analysis are another example of areas for improvement. “Early on we made quite a large investment in AI, as it’s now known,” Brown said, starting with hidden Markov models, but quickly shifting to discriminative neural networks. Now, he said, Nanpore has taken advantage of the AI explosion and integrated new technologies into the nanopore sequencing application.  

“We’re now using transformer-based models, state-of-the-art models, for high-end nanopore basecalling,” Brown said. These SUP transformer models are enabling even higher accuracy; V5 SUP, the version currently available in Dorado 0.7, achieves Q26. Q30 simplex accuracy is available, Brown said, with some tradeoffs. “You lose about 15% of yield at the moment, and the basecaller is four times slower.”  

“We’re very good in coding regions,” Brown said. “It might surprise you that we’re doing very well in the clincal world. A lot of clinical researchers like Nanopore because it’s good enough—because they’re mostly looking at coding regions and the indel thing doesn’t really hit them.” 

T2T Tomorrow 

Brown is known for “setting ridiculous targets,” he said, as he transitioned his presentation to Oxford Nanopore’s vision of the future. His first ridiculous target: a Nanopore-only, human telomere-to-telomere (T2T) assembly. With four flow cells Brown believes Oxford Nanopore has achieved a T2T assembly: two for ultra long reads, one for 6B4 for consensus polishing to better estimate homopolymer length, and one for Pore-C (a Nanopore-specific version of Hi-C).   

The human T2T genome product bundle—a set for six human genomes, or 24 flow cells plus prep and wash kits—will be priced at $24,343 and will begin shipping in July, Brown reported. The analysis pipeline is also bundled and automated in Dorado.  

“That is the first step toward ‘Nanopore is all you need.’ That’s the idea,” he said.  

But he has bigger visions than just a T2T assembly; Brown proposes sequencing an entire chromosome as a single molecule. Could you skip the 6B4 step and resolve the homopolymers some simpler way? And Pore-C is quite a long workflow and complicated, he added. Could you make simplex reads that were good enough and long enough to just use those?  

“With Nanopore sequencing, if you can get a stupidly long molecule to the pore, the pore will—most of the time, unless there’s a crosslink or something—process that fragment. It just goes through,” he said, adding later that histones don’t even bother the pore. “It’s a really big differentiating feature of this platform. You don’t need to break things up or amplify them; the read length isn’t limited by the sequencing chemistry per se.”  

There are still challenges, and Brown listed five: you need an unconventional extraction process that doesn’t break the DNA; you need an adaptation method; you need a delivery mechanism that doesn’t involve pipetting or diffusion; and you have to physically transport this megabase-long piece to the pore. Finally, he pointed out that sequencing the largest human chromosome would take about 78 hours currently and the pore itself needs to live that long.  

In the past six weeks, the Nanopore team has been able to sequence yeast chromosomes as a single molecule, and he expects further optimization of each of the challenges will enable them to, “move on to a bigger organism with longer chromosomes.” 

There are other iterative changes Brown says Nanopore is making to improve output. New buffer chemistry within the flow cells can raise yields by 75% without having to refill flow cells. Nanopore is using AI-guided motor engineering to improve motor speed and maintain accuracy. And longer nanopores can give improved homopolymer accuracy. All of these changes drive toward simplex-only chemistries.  

Beyond Canonical DNA 

Another area where Nanopore has historically performed well and Brown expects further investment and improvement is calling non-canonical bases. In the research environment, Nanopore can call all types of non-canonical DNA including 5mC, 5hmC, 6mA, and 4mC. Coming next, Brown said, is detecting genomic uracil, and next on the hit list are 8oxoG and Ribo-bases and other DNA damage. 

Brown acknowledged that we don’t quite know what these markers may mean, or what we’ll learn by tracking them, “but they’re all detectable on Nanopore. We just have to fold them into the basecallers.”  

He added that the absence of believable ground truth for these non-canonical bases and testable DNA damage means that Oxford Nanopore has invested a “ton of work developing synthetic test constructs and synthetic libraries using a lot of fancy chemistry to make believable training sets and test sets to validate these things.”   

Oxford Nanopore is also the only method for directly sequencing native RNA, Brown said. The RNA004 chemistry was released last year. “Entertainingly, there was a bug, I suppose, in our software that meant that we weren’t writing out short RNA molecules. It turns out, if you fix that bug, all the short, natural RNA molecules appear. You can actually get 120 million of them on a PromethION flow cell.”  There’s an application here for liquid biopsy, Brown said.  

Product Updates 

Brown closed his keynote giving software and hardware updates. Dorado 0.7.0 is available on GitHub, with updates to basecalling, read correction, modified bases, poly(a) tail estimation, barcode demultiplexing, and faster alignment. EPI2ME integrated analysis pipelines are currently available on desktop, but Brown said updates were coming soon for cloud compute, device integration, and integration with analysis providers.  

Brown announced early access to the latest MinION, the Mk1D, which is USB-C compatible, has temperature controls, and simpler installation. Early access devices will ship in July he said, with broader access in September.  

“We gambled a while back that the Apple processor line—built to do machine learning—things on it would scale at low power, and that’s what’s happening.” The Mk1D works with the Apple iPad with an M4 GPU and 5G mobile data connectivity. The system will be very field ready, Brown added.  

He also announced a new flow cell design to make loading simple and robust. “Finally, we’re going to revise the MinION flow cell design with a much simpler input method, much more user tolerant… and it deals with goopy samples, so it’s good for megalong samples.” Brown says to look for revised flow cell prototypes in late in 2024.  

The next new product he announced “is not one of mine, actually”. ElysION (formerly TurBOT) is a standalone robotic nanopore sequencing device with automated workflows, multiple sequencers, integrated compute, and data processing on board.  

“It fits a niche, really,” he said. “It’s a very hands-off, biopharma QC-oriented, lab automation-oriented, high throughput testing lab-oriented device. A departure, for us, really. But a happy departure! I’m totally cool with it.”  

Brown quoted a list price of $270,000 and said ElysION has a “clear pathway” for regulatory approval. “Don’t be surprised! Nanopore is popular with… clinical testers because the mutation detection is actually very good.”  

On the opposite end of the user spectrum, Brown also introduced TraxION, a factory-loaded flow cell integrated with a sequencer that takes samples via swabs, tips, or pipettes. It’s designed to have a single point of use and to be very simple to run individual samples. “It does the end-to-end sample prep (if there is one) and it will also run the sample as a sequencer.”   

Finally, as always, Brown gave updates on the ASIC—the application-specific integrated circuit—that is connected to the sensor and measures current flow across the nanopore. The newest ASIC is very low power to run on portable devices, has 400 channels, and is so cheap—“I can tell you this!”—that it’s disposable, Brown said. The sequencing is as good as, or even slightly better, than current ASICs, Brown said.  

The new ASIC will feature in a family of new instruments including the MKII MinION and a multiple flow cell MKII instrument so that each flow cell functions a bit like its own lane and up to 8 can be swapped out.  

But Brown also looked ahead, teasing an ASIC with a field-effect transistor design and 100,000 channels. It’s the same chemistry but with a radically different chip design. Using a voltage sensor will let Nanopore increase the well density by about 10x. Prototypes are already in flow cells, and he predicted a benchtop instrument using the new ASIC in 2025 or beyond.