NEW YORK – BGI is entering new territory as the group’s researchers unveiled their first nanopore sequencer in a preprint this week.
The Chinese omics conglomerate has granted the global rights to commercialize and distribute the technology to affiliate company MGI Tech, which is staying mum for now about its commercial roadmap.
Xun Xu, director of BGI Research, a unit of BGI Group, said his organization sees a lot of benefits in developing a nanopore sequencing platform. "It is also very complementary to the short-read [technology we have]," he said. Xu's team described its nanopore sequencing technology in a preprint in BioRxiv this week.
Named CycloneSeq, BGI's platform is similar in principle to that of Oxford Nanopore Technologies. Both technologies use helicase motor proteins and protein nanopores to read DNA molecules, generating current signals that can be translated into sequences using deep learning models. However, Xu noted that BGI, which has been developing its nanopore technology for a decade, has put its own spin on the protein engineering, chip design, sequencing chemistry, and basecalling algorithm.
More specifically, Xu said the motor protein and pore deployed in CycloneSeq were evolved from bacterial proteins derived from samples collected in BGI’s previous deep sea expeditions. "We are continuously putting efforts into collecting samples from extreme environments all over the world," he said, including an expedition to the Mariana Trench in 2021.
In their paper, the BGI researchers noted that their motor protein, a helicase that unwinds double-stranded DNA, can achieve a speed of approximately 380 bp per second. Meanwhile, their pore protein can translocate single-stranded DNA and generate currents "with high signal complexity and good signal-to-noise ratio," according to the preprint.
The sensor chip features a high-density microwell array that holds the nanopores. Each microwell and sensor unit is spaced approximately 200 µm apart, and sensor units are arranged in a honeycomb pattern to maximize density and enhance sequencing throughput.
Moreover, the BGI team designed the microwells to have small diameters, boosting the signal-to-noise ratio and improving sequencing accuracy.
To improve data quality, Xu said BGI developed so-called nanopore local chemistry (NCL) sequencing, where the chemical environment on each side of the nanopore is asymmetric. It involves the controlled release of magnesium ions, which are embedded within the flow cell rather than supplied with the sequencing reagents. After applying the transmembrane potential, magnesium ions are translocated through the nanopore, maximizing their concentration near the pore. Together with ATP, the magnesium ions fuel the unwinding and translocation of DNA molecules through the nanopore.
Lastly, Xu said the team developed a basecalling algorithm trained with deep learning models using unlabeled data, which were then fine-tuned for enhanced accuracy. The CycloneSeq sequencer can also perform sequencing and base calling at the same time, according to the paper.
According to BGI, when tested with E. coli DNA, CycloneSeq cumulatively generated more than 50 Gb of data with 107 hours of sequencing on a single flow cell. For human genomic DNA, the platform was able to produce long reads with an N50 of 33.6 kb, while achieving a raw read accuracy of 97 percent.
"If everything works as it is written in the paper, then it is pretty cool," said Christopher Vollmers, a professor at the University of California, Santa Cruz and a nanopore sequencing expert. "It would be behind where [Oxford Nanopore] is today by a few years in terms of throughput and accuracy, but it's not way off."
Vollmers pointed out that BGI has not yet made publicly available important information such as sequencing data, raw nanopore signals, as well as the basecalling code and its training dataset, making it impossible for the community to independently verify the technology’s performance. "There is a whole bunch of things missing from this paper," he said. "As it stands, this is not a real paper … this would have been a great white paper."
Kin Fai Au, a professor at the University of Michigan and a long-time nanopore sequencing user, said he feels "positive" and "excited" about BGI's nanopore sequencing technology overall. Still, based on his experience, there tends to be a discrepancy between the advertised metrics from sequencing companies and the technology’s real-world performance. Therefore, external benchmarking is still necessary to evaluate CycloneSeq's true performance. "Some of the figures [in the study] are too good for me, but we will wait for the third-party evaluation," he noted.
BGI did not disclose the format of a commercial nanopore sequencer in its paper. Xu said there will likely be a few platforms developed using the CycloneSeq technology, with one of the first being a "shoebox-sized" instrument that can accommodate two flow cells.
According to Xu, BGI Group, formally named BGI Shenzhen, has established a new subsidiary called Hangzhou BGI Xufeng Technology that holds the IP related to CycloneSeq. Earlier this year, BGI Xufeng granted the exclusive rights to commercialize and distribute CycloneSeq globally to BGI Group-affiliate MGI Tech.
"By distributing nanopore sequencing products, MGI will be able to provide users with a comprehensive long-read and short-read sequencing tool, not only facilitating the expansion of long-read sequencing equipment for existing customers but also helping to broaden downstream sequencing applications," an MGI spokesperson said in an email.
She declined to answer specific questions regarding CycloneSeq's commercial roadmap at this time, including its price and availability, noting that the company will disclose its plans for CycloneSeq at an official launch in early September.
It also remains unclear whether BGI's nanopore patents overlap with those of Oxford Nanopore, which is widely perceived to have a tight grip on nanopore sequencing IP. Both Xu and Oxford Nanopore declined to comment on the subject.
Even if CycloneSeq proves its performance, BGI and MGI will likely have much legwork to do to drive the technology's adoption in the US, especially against the backdrop of a deteriorating US-China relationship and looming legislation that proposes to target both entities’ products and services.
"It's cool that it exists, but I'll probably never use it — not because I wouldn't want to, but because I probably won't have the opportunity," Vollmers said. "That's the truth with the US-China relationship."
"The US may not allow BGI to sell their service or their machines in the country," Au said. "In that case, we have no way to utilize this technology."
Even within China, BGI will face domestic competition from the likes of Qitan Tech, a nanopore sequencing startup that purportedly launched a handheld, low-throughput platform in 2021.
Some researchers believe the emergence of new players in the nanopore sequencing market, including a platform Roche reportedly has been working on for years, might move Oxford Nanopore to re-assess its commercial strategy.
According to Scott Tighe, technical director of the University of Vermont Advanced Genomics Lab, Oxford Nanopore increased instrument prices a lot lately, and "this might cause them to think twice." The company "forces users to buy their instruments bundled with reagents and flow cells," he said in an email. "This is not a good model and is quite upsetting for potential [Oxford Nanopore] users since we often buy instruments separate from consumables."
Even so, Tighe does not think BGI, which is "11 years behind in nanopore, and in no position to gain any market share," will pose a real threat to Oxford Nanopore at this point.
"I think once they get a peer-reviewed paper out and a few hundred units out to people, maybe in five years we’ll learn more," Tighe said. "Until then, I don’t think Oxford has any concerns."