NEW YORK (GenomeWeb) – Oxford Nanopore Technologies plans to release several updates to its MinIon sequencer next month, including a new nanopore version and motor protein that will increase sequencing speed, yield, and accuracy, as well as improved base calling software.
Over the next few weeks and months, the company also plans to ship direct RNA sequencing kits to a group of developers and VolTrax automated library prep devices to early-access users. It also plans to deliver the first flow cells to early customers of its PromethIon high-throughput sequencer.
Longer term, company researchers have been working on a sensor with a million channels that will enable very high-throughput sequencing, solid-state nanopores, a DNA enrichment method that uses a modified Cas9 enzyme, a combined sample and library prep device, a small nanopore sensing device called SmidgIon that plugs into a smartphone, and an adapter to the MinIon that makes it compatible with small and inexpensive flow cells.
Clive Brown, Oxford Nanopore's chief technology officer, talked about these plans and projects during a live webcast yesterday.
Next month, the company will release a new MinIon flow cell and sequencing chemistry with a new pore version and a new helicase motor enzyme that will come with an updated base caller and MinKnow software. Customers will be able to order these updates on Oct. 10 and they will start shipping Oct. 17.
The new pore, version R9.4, is more stable over time than the current R9 pore, leading to higher yields per run, and has greater accuracy for 1D reads. The new motor enzyme, E8, operates at a speed of 450 bases per second instead of 250 bases per second, almost doubling throughput with no significant drop in accuracy. The firm also changed the stalling chemistry that keeps the motor protein still before the DNA enters the pore, which had been optimized for the R7 pore and did not work well for the R9 pore. In addition, the new flow cell will load the DNA on beads, further increasing the yield per pore.
In Oxford Nanopore's hands, using these updates, the MinIon now generates about 9 gigabases of data within 48 hours, Brown said. An output of 10 gigabases won't be far away, and further improvements up to 20 gigabases seem possible, he added. Ultimately, the sequencing speed could be increased to 1,000 bases per second. About 90 percent of internal runs currently yield more than 5 gigabases per run, he said.
With the R9.4 pore, the base calling accuracy is now greater than 90 percent for 1D reads and about 97 percent for 2D reads when run at 250 bases per second, and around 92 percent for 1D reads when run at 450 bases per second. 2D reads are currently not available at the higher speed, and Brown said the company is trying to make 1D reads the de facto standard going forward.
The company will release two sequencing kits: a 1D ligation sequencing kit running at 450 bases per second that promises 10 gigabases per flow cell, and a 2D ligation sequencing kit that runs at 250 bases per second and has a 2.5 gigabase yield for 2D reads and a 5 gigabase yield for 1D reads.
The price of a flow cell is $500 and library prep costs $100 per sample, so the cost per gigabase for a single sample will be $60 for flow cells yielding 10 gigabases and $120 for flow cells generating 5 gigabases of data, which Brown said is comparable to some short-read sequencing systems.
Also, up to six libraries can be run on a flow cell sequentially, reducing the cost per sample to about $200, and up to 12 samples can be barcoded, bringing the cost per sample to about $50.
Some but not all of the enhancements have already been available to a select group of developers, who have had yields of 4 to 5 gigabases per run, Brown said.
In terms of read length, Oxford Nanopore's internal record remains 255 kilobases for a single 2D read or 510 kilobases for a single 1D read, and Brown said a megabase read should be possible with bead-based sample prep that treats the DNA more gently.
The company has already made another upgrade to the MinIon flow cells, called SpotOn, that "nobody noticed," Brown said, switching the plastic material, which increased the shelf life and improved reliability by preventing air from seeping into the flow cells, and adding an input port directly above the nanopore array. SpotOn flow cells can be shipped at ambient temperatures, while the old ones had to be kept cool, he said.
About 3,000 MinIon sequencers are currently in customers' hands, Brown said. He noted that customers who upgrade to the R9 flow cell will be able to take advantage of any future upgrades, while customers still running R7 flow cells cannot use those upgrades. Should the company switch pores again in the future, which it does not currently plan to do, it will still sell the R9 pore in parallel, he said.
With regard to base calling, the company recently moved from a hidden Markov model to a recurrent neural network, which can take more information along the pore into consideration and has led to increased accuracy. There is still room for improvement on the algorithmic side, Brown said, in particular with regard to training the algorithm and the quality of the training set.
The main type of error in Oxford Nanopore's reads is still related to homopolymers, which the current base caller rounds down to five bases. The company is working on base callers that can call longer homopolymers accurately by exploiting the time-dependent signal in the ion current, Brown said. In addition, it may make changes to the sequencing chemistry that improve homopolymer reading.
Another improvement the company wants to make is to enable its base caller to read base modifications. These are already present in the ion current signal, Brown said, but the base caller is currently trained to read only the four canonical bases. This also means that the base caller tends to make errors when a damaged or modified base is present in the DNA.
Three base calling options are currently available. In August, the company released a real-time local base caller as part of its MinKnow software, which currently only supports 1D calling but will soon be available for 2D calling. It is also working on hardware acceleration for the base calling process. Secondly, the firm released the source code for Nanonet, an offline local base caller for which it provides limited support. It is currently developing a new package called Albacore, which it plans to make available to customers and developers on Oct. 17. Future innovations to base calling will be available through the Albacore pipeline, Brown said. Finally, cloud-based base calling is still available but will be phased out by early November.
In addition to the "core upgrades" to the MinIon in mid-October, the company plans to release additional kits: an updated flow cell wash kit and an updated rapid 1D kit in October, a short fragment kit and a rapid PCR kit in November, and a PCR-based barcode kit and a rapid amplicon kit in December. The rapid 1D kit, which uses a transposase to fragment the DNA and add adapters, will be quicker as well as more robust to "dirty samples" than the previous one, Brown pointed out, and may use lyophilized reagents.
PromethIon in the starting block
Brown also provided an update on the PromethIon early-access program, which will likely be closed to new members by the end of this year. A handful of PromethIon sequencers have now been shipped to early-access customers in Europe and the US. The instrument, which takes about an hour to install, consists of two main modules, a sensing module on top, which contains data collection electronics and flow cells, and a compute module on the bottom. Early-access users will likely receive a new compute module in the near future that enables online base calling, Brown said.
Flow cells for the PromethIon are not ready yet. The company can already generate sequence data of similar quality to the MinIon on PromethIon flow cells, but it is working on improving their yield by increasing the percentage of available pores, which is mostly related to the manufacturing process, Brown said. The flow cells "are within weeks of being available," and the company plans to ship them once it has about 10 to 12 PromethIon instruments installed at customer sites.
Each PromethIon can take up to 48 flow cells, each with up to 3,000 nanopore channels and four sample inputs, that can be run independently. At a sequencing speed of 250 bases per second, each flow cell can theoretically generate up to about 130 gigabases in 48 hours, and all 48 flow cells may be able to yield up to 6.4 terabases per 48-hour run. These yields may double with increased sequencing speeds.
Cas9 for target enrichment
Responding to an increased demand for sequencing-specific DNA targets in a complex sample, Oxford Nanopore has also been developing an enrichment technology that makes use of a disabled Cas9 nuclease and associated guide RNA probes that targets a 20-mer region in the DNA sample. Cas9 and the guide RNAs bind to the regions of interest and the DNA is prepared for sequencing using a 1D library prep kit. Only DNA molecules that have Cas9 bound will stall the motor protein and be available for sequencing. When the DNA is trapped on the pore, the Cas9 complex is kicked off and sequencing can begin. A tether attached to the Cas9 enzyme can be used for additional enrichment.
With a different nanopore, the company can alternatively just detect Cas9-bound DNA fragments without sequencing them, a kind of counting application for quantifying specific sequences in a sample. This application is very fast, enabling the measurement of millions of molecules in a few seconds, Brown said.
There is still work to do on increasing the specificity of the Cas9 enrichment approach, he said, and the sensitivity could be further enhanced by using beads or different tethering strategies. Oxford Nanopore is "quite likely" to make the enrichment method available in kit form "within some reasonable timeframe," he said, though researchers could also develop their own enrichment schemes using disabled Cas9.
Designing the targeting probes will be the most important part of those kits. For "well-known tests," Oxford Nanopore may supply standard sets of Cas9 probes, he said, which would allow the company to deploy rapid tests on mobile nanopore devices.
A million channels, SmidgIon, Zumbador, VolTrax, Flongle
Oxford Nanopore has been working on a sensor with a million channels that uses a field effect transistor (FET)-based technology to measure the resistance through the pore, which can be a protein pore or a solid state pore. The company has achieved proof-of-concept sequencing on this sensor using the R9 chemistry, Brown said, noting that the signal-to-noise ratio of the sensor is very good, and the read-out speed is high. The new sensor will require a new application-specific integrated circuit (ASIC) and chip platform, which are currently in development. With a million channels and an enzyme speed of 1,000 bases per second, the throughput of this system could reach a gigabase per second. "This is our next-next big technology shift" and is at least two years away, Brown said.
The company has also made "significant breakthroughs" in the development of solid-state nanopores, he said. It is now able to measure DNA passing through solid-state pores, controlling the translocation speed with a helicase enzyme, and has obtained sequence-specific signals. It is not clear yet whether solid-state pores will be superior to protein pores in terms of the signal quality, Brown said, but they may be more robust and more practical for other reasons.
The SmidgIon, a portable nanopore sensing device for rapid and frequent testing that the company first mentioned earlier this year, will be available in late 2017. The device is smaller than the MinIon but uses the same underlying technology and plugs into a smartphone. Most of the electronics are inside the sensor box and the flow cell is small and inexpensive. The device, which will have either 128 or 256 nanopore channels, will support all sequencing chemistries for the MinIon and PromethIon, Brown said, and generate up to 400 megabases of data per hour. Together with the Zumbador automated sample prep device that is in development and the target enrichment technology, SmidgIon could be used for field testing, he added.
Oxford Nanopore continues to work on Zumbador, which would allow users to use the MinIon in the field outside a lab. The device will take liquid samples such as blood, saliva, or food, extract DNA from it, and prepare a sequencing library. Cell lysis, which is still under development, will happen in liquid phase in the top part of the device. The bottom part will carry out library prep, using rapid 1D kit reagents, and will contain dried DNA capture beads, lyophilized reagents, and, possibly, targeting reagents. The beads carrying the DNA library will then drop into the flow cell, driven by gravity. "The bit closest to the flow cell is by far the most mature," though the efficiency is currently not high, Brown said, noting that there is no release date for Zumbador yet.
In conjunction with the development of the SmidgeIon, the company has also developed an adapter for the MinIon, called "flow cell dongle" or Flongle, that contains all of the sensing electronics. Using this adapter, the MinIon can be loaded with inexpensive plastic flow cells containing the pores with 128 or 256 channels. Oxford Nanopore plans to commercialize the MinIon adapter and small flow cells, but they will not be available until at least six months from now, Brown said.
Work on the VolTrax, which will allow for automated library prep, is also continuing. The small instrument will be powered by a USB port and be able to heat, cool, exert magnetic forces, and move droplets. It takes disposable flow cells with multiple sample and reagent ports. While the developer version will require users to pipette reagents, the company plans to sell flow cells preloaded with reagents for specific workflows, including library prep protocols, cell lysis, and DNA extraction.
VolTrax still requires some amount of pipetting, but in the future it might be possible to attach it to a pipette-free Zumbador device, Brown said. Registration for the VolTrax early-access program will open Oct. 11 and the program will likely start in early December.