The Oxford Nanopore Technologies (ONT) tech update that occurred on Thursday generated significant excitement among scientists interested in Next-Generation Sequencing (NGS). Many researchers had been eagerly awaiting this update and were particularly interested in the advancements made by ONT.
The overall conclusion drawn from the update is that ONT has now established itself as a Q30+ NGS company. This means that their sequencing accuracy has reached a high level of modal 99.9% accuracy or higher, comparable to most short-read technologies in the field (Illumina, Element Bio, MGI Tech), and the new PacBio Revio instrument, which generates 90-100Gb of Q20-Q30 data at a 15-20kb range (the claim is that it’s “a median of Q30 or better”). This achievement is crucial because high sequencing accuracy is essential for obtaining reliable and accurate results with the software that the NGS community has gotten used to in the last 20 years.
In addition to achieving high accuracy, ONT's update highlighted several advantages over other NGS technologies. One of the notable advantages is the production of ultra-long reads, which refers to the ability to sequence very long single-molecule DNA fragments in a single readout. The current record stands at 4 megabases. This is particularly beneficial for applications such as genome assembly, structural variant detection, and haplotype phasing, as demonstrated lately by the results of the Telomere-to-telomere project (T2T).
Another area of interest is ONT's base calling capabilities for modified bases, specifically 5-methylcytosine (5mC) and importantly also 5-hydroxymethylcytosine (5hmC). These modified bases play essential roles in epigenetic regulation and can provide valuable insights into gene expression and regulation patterns, but crucially, they have been demonstrated to have a big impact in the multi-cancer early detection methods (MCEDs) based on assaying cell-free DNA from liquid biopsy blood samples.
ONT also introduced adaptive sampling for the P2 Solo instrument, a feature that allows for dynamic adjustment of sequencing parameters during a run. This adaptive approach optimizes sequencing resources and enables more efficient data generation by focusing on regions of interest or increasing coverage in specific genomic regions.
An important aspect emphasized in the update is the low capital expenditure (CAPEX) and portability of ONT's instruments, of which more were announced at LC2023. This means that ONT's sequencing devices are orders of magnitude more affordable compared to other near equivalent technologies, and they offer the flexibility to be used in various settings, including field-based research or resource-limited environments.
As evidence of their progress, ONT announced the shipment of over 400 P2 Solo instruments within less than two quarters. This indicates a significant adoption of their technology and demonstrates the accessibility of their PromethION flowcells to a large number of customers.
The presentation started with a walk through the history of the duplex sequencing, and how the hairpin method was abandoned to substitute it for a method were each strand is encouraged to go through a pore by modifying the adapters, the point at which the duplex rate reached 30%.
An improved version followed which generated close to 80-85% pairing. In the slide shown below, they describe the results of running this new method by giving the library prep and flowcell loading task to 9 different graduate students. Eight out of 9 of the beta-testers reached the desired level of duplex calling in the sequencing runs.
This new system is called “High-Duplex” or “Super-duplex”, and with the latest version of the PromethION flowcells comes labelled as R10.4HD. The basecalling uses methods similar to those used in Machine Learning applications such as ChatGPT.
The duplex data makes a big difference to applications like assembly but also variant calling, although a big improvement was already seen in H2 2022 with the introduction of Kit14.
Duplex calling is agnostic of the read length, and this is one of the advantages of the ONT technology that I will emphasize later on behind the curtains, in the market positioning part of the post.
The two crucial performance values from the tech update were the following:
R10.4HD renders 60Gb of duplex data from a PromethION flowcell for PCR-free human DNA samples, but crucially,
R10.4HD renders >100Gb of duplex data (plus about 50Gb of simplex) when the DNA samples has been pre-amplified with a PCR-based library prep.
The second number, rounded up to 100Gb of duplex data from a PromethION flowcell, is the crucial number to keep in mind here. I’ll draw comparisons to the competitors later on in the post.
ONT has a route to even higher duplex rates, with improvements enabled by the more sensitive way of sensing the signal in the recent tech updates.
In true Apple style, the first part of the talk had always a “released now in store” slide for each of the announcements, including R10.4HD.
Changing gears to Simplex, the awaited announcement of the new 5kHz update means that now the basecalling of HAC reaches the same levels of accuracy of SUP but with 1/10th of the compute requirements.
An example of this can be seen on the performance based on an NVDA 0.00 Nvidia A100 GPU system using the new de facto standard base caller, Dorado.
This is an update from the current 4kHz system, and with the 5kHz system, it unlocks further capabilities that were explained later on.
Taking the 5kHz system as a baseline, ONT sees a doubling of the throughput potential across the board with improvements in several fronts, aiming at bringing this updates in the next 12 months.
The P2 Solo and now the P2 instruments have made a big splash since they started selling, with more than 400 P2 Solos out in the field. Adaptive sampling is now enabled and fully automated in the P2 instruments.
There was an update on software as well, and the EPI2ME platform. There are three different versions, or layers, of EPI2ME, namely the “OG”, EPI2ME Labs and EPI2ME Github. Users of different levels of expertise can interact at the layer they feel comfortable. The NextFlow community will find it easy to use, and more importantly, to contribute to.
Lots of work in turning command-line workflows into clickable applications that can lower the barrier of entry to clinical assays.
The DNA modifications base-calling part of the tech update was equally impressive: 5mC and 5hmC have a big role to play in MCED in the next few years, but ONT aims at an “All context” base-calling now.
We have started seeing papers published detailing the DNA damage modifications seen in cfDNA and tumour DNA samples, and how those can further inform cancer diagnosis as well as other diseases, such as neurodegenerative diseases such as Alzheimer’s and Parkinson’s. ONT is very well positioned to play a role in unlocking this new area of science and diagnostics.
Other DNA modifications have an importance in field such as plant science. In many cases, we see plant science lagging behind human health in the tech available to them, so ONT is offering solutions to solve scientific problems that are segment specific.
Another important update is the “Mods to break up homopolymers - 8D4” slide. This may have not seen too relevant to people, but it is possibly the last wall that ONT will bring down to finally convince the sceptics of the last two decades about the ONT technology future.
Some of us will see traces of the “Illumina Complete Long Reads” solution in this idea, which we haven’t been told it was the Longas technology yet (but, really, it was). Since adding salts to a PCR solution is *not patentable*, Illumina has no means to stop ONT coming up with a method that finally resolves the homopolymer calling with nanopore sequencing. No more excuses for calling ONT reads Q30+, in all contexts.
Another important update, following the same thinking as in DNA mods, is the RNA004 upgrade.
This now unlocks the science of RNA mods, of which there are more than 100 known, to all interested in studying.
If that sound like a very niche part of Science that shouldn’t matter to most, the following slide showed an example RNA mod that is actually very close to all of us, at least for people who took an RNA-based COVID19 vaccine in the last 2-3 years. The reason being that RNA vaccines are produced by introducing a mod to the RNA bases that increases their expression once injected, adding more antigen available for our immune system to react against, in this case, SARS-CoV2 virus when infected.
Vaccine producers are I am sure very glad to have technology like ONT to QC their vaccine manufacturing, which will continue to be crucial as the world gets used to monitor and react to future infectious diseases such as COVID19.
We’ve gotten used to the line-up of instrument products getting busier, here with the P2 and P2 Solo adding to the family.
The P2 is now shipping, with integrated compute, easy to use, all-in-one solution for PromethION flowcells.
On the instrument side, the new MinION, together with a docking station to attach to an Apple iPad, was announced.
Given that the new iPads and MacBooks have Apple Silicon chips, a version of the ARM CPU+GPU chips we have in our smartphones, it means the power of this compute can keep up with base-calling the data coming out of the new MinION MkID.
At this point, the presentation shifted gears again, and went into describing what is ongoing work in the company.
Work ongoing to increase the overall throughput is now nearing release. Increasing outputs: motor work, some up to 1000bp with high accuracy, at 5kHz will mean more data at same or higher accuracy.
Buffer development on cis and trans side, can have important effect. Secondary structure formation, G-quads on the trans side, deblocks better. It explains why genomes with low-complexity regions, like chicken and some fish (Takifugu is a famous one), now much better output.
There are samples that are very precious to customers. The ability to harvest and reload the sample as many times as possible is unique to ONT technology. This included re-prepping the sample. It means that samples can be biobank’ed forever.
Perhaps the most dramatic announcement of this tech update was the ongoing work with the new ASICs. Even though we think of ONT as a DNA-sequencing company, a large part of their expertise resides in the semiconductor design.
The new ASIC is a different architecture, drawing parallels to computing parlance. A single MUX, i.e. one corridor with a door at the end, compared to the previous iteration of ASICs which had 4 doors at the end, where only one door opens at a time. New ASICs are very low power, suitable for portable devices with small battery requirements.
Based on the new capabilities of this new ASIC family, ONT will release in the future a next version of their MinION, called MkII.
This will give an estimate of 5-10Gb per flowcell, and it can be configured in a grid in a similar fashion to the GridION instrument, only here, a laptop-level computer will be able to keep up with the base-calling of several of this SmidgION flowcells.
The equivalent of the new ASIC on the MinION flowcells will give more data in a shorter period of time than the current version.
This new ASIC, which is very cheap to produce, offers all sort of options for ONT. One of them is to have an integrated step of sample preparation plus flowcell loading, in a revamped VolTRAX device that can load directly into the new SmidgION flowcell.
This means that the only step required for the user is to load the DNA or RNA sample on this TraxION device, and have it perform all the steps towards sequencing the sample without user intervention.
The tech update completed with a calendar chart of the releases coming in 2023, and a call for people to attend the next two updates, in Singapore in September and in Houston in December.
How does this put ONT with respect to their rivals in NGS?
Behind the curtain, I’ll explain the consequences of the ONT product updates, their ongoing work with the new ASICs and how this fits into the crowded field of NGS in H2 2023 and beyond…