This story has been updated to correct the number of Element sequencers in the field and include additional information from the company; to clarify the Element flow cell output reported by the Broad Institute; and to clarify some specs of PacBio's SBB sequencer.
ORLANDO, Florida – Get a couple hundred scientists in the same room and it's hard to get everyone to agree. If there's one truism at the Advances in Genome Biology and Technology annual meeting, though, it's that sequencing costs, as far as they've come down over the decades, are still a barrier to crucial studies, from population genomics to single-cell experiments to cancer research. As a corollary, most would agree that any competition for Illumina, still the dominant market force in sequencing to date, is good.
With nearly half a dozen new short-read sequencing instruments slated for release in the US over the next year or two, it appears that competition is on its way. But in the absence of much user experience, you won't find any consensus on those instruments among researchers.
Element Biosciences, which unveiled its mid-throughput Aviti system in March, and Ultima Genomics — only a rumor for many until its first announcement last week — were the NGS companies with the highest profile at the conference, the benefit of being in the second-highest tier of sponsors.
Dozens of invited and sponsored talks, posters, and marketing displays featured Element's mid-throughput Aviti and Ultima's high-throughput UG100. Researchers at the Broad Institute have completed beta testing runs for both platforms and appear set to keep them on. Other researchers presented data on key applications of those platforms, including single-cell sequencing, cell-free DNA sequencing, and Hi-C-based epigenetics assays.
Also trying to elbow their way into the NGS market are Singular Genomics' G4 and, in August, MGI Tech's DNBSeq-G400, both mid-throughput instruments. Further, Pacific Biosciences is crossing over into short-read sequencing territory with a sequencing-by-binding technology acquired from Omniome last year.
Reactions to each of these instruments were mixed and, in some cases, in flux. Ultima provoked the widest range of responses, from the emphatic to the diplomatic.
In a Wednesday lunchtime talk sponsored by Ultima, following a talk about new Hi-C genome interaction methods developed with Ultima's sequencing in mind, Erez Lieberman Aiden of Baylor College of Medicine said this year's meeting was reminiscent of AGBTs from the mid to late 2000s, when NGS first came on the scene. "It felt like walls were coming down; things that had been inconceivable were now straightforward," he said. "I think some of us have missed that heady feeling. My impression is, it's back."
Certainly, Ultima's claims of $1 costs per Gb of sequencing data is fueling that feeling. But Ultima's use of off-instrument emulsion PCR to generate amplified DNA on beads ready for sequencing may keep some customers at a distance.
"It scares me to see this," said Renata Pellegrino da silva, scientific director of core facilities at the Children's Hospital of Philadelphia Research Institute. "I'm having flashbacks to the SOLiD sequencer," a now defunct sequencing platform by Life Technologies (now part of Thermo Fisher Scientific) that also employed off-instrument emulsion PCR to generate amplified DNA. "I hate that they're a monopoly, but Illumina did a great job to put everything in one instrument," she said.
Element's offering also impressed many. "Quality looks extraordinarily good all the way out to 150 nucleotides," said Chris Mason, a sequencing technology expert and professor at Weill Cornell Medicine, who presented on his use of the Aviti platform in single-cell studies of astronauts. Single-cell ATAC-seq (assay for transposase-accessible chromatin by sequencing) data were very consistent, "almost indistinguishable" from Illumina, he said. "In some cases, more high-quality fragments are aligning to the genome."
But some questioned whether the newcomers had shown enough to precipitate a large-scale shift from Illumina's well-proven offerings. "To be blunt, if Illumina was walking in and presenting one of these sequencers, I would have way more confidence that we knew it was going to be there and it was going to deliver," said Tom Wilson, director of the University of Michigan's Advanced Genomics Core. Ultima, Element, and Singular "are basically brand-new players in the field … it's not to criticize them, it's just a reality."
Illumina, which just a few years ago controlled 90 percent of the US sequencing market, according to the US Federal Trade Commission, has not been entirely quiet, either. On Tuesday, the firm announced a new 2x300 bp paired-end sequencing kit for its mid-throughput NextSeq 1000 and 2000 instruments, a run mode previously available only on the low-throughput MiSeq. But the company appears to be willing to cede the spotlight for now, although it has promised more information about its next-generation “Chemistry X” that it claims will lower both costs and run time. The firm also hosted talks in its hospitality suite touting the so-called Infinity synthetic long read product and improvements to its Dragen bioinformatics suite.
Illumina also disclosed that it recently acquired Emedgene, a genome interpretation company it partnered with on rare disease-causing variant detection in late 2020. Financial details of the deal were not disclosed.
MGI Tech, the other sequencing company with more than just a handful of instruments out in the field, albeit not in the US or in Europe, announced at the dawn of the conference that it will take another stab at the US NGS market this August, as soon as Illumina's IP shackles come off.
The company plans to launch its DNBSeq-G400C, a version of the DNBSeq-G400 sequencer that is based on its newer, antibody-based sequencing chemistry, CoolMPS, but it stressed that it will not accept any orders before Aug. 29. Unlike in previous years at AGBT, MGI Tech mostly kept a low profile and did not present new data of its sequencing platform at the conference.
Regarding the sequencing startups, their underlying technology was of great interest to researchers.
Ultima's use of emulsion PCR means the company requires a separate machine for sample prep. The instrument is already large — bigger than a household refrigerator — and the sample prep station is about that size, as well. The towering UG100 was on display in the Ultima suite; however, the sample prep station was not.
The emulsion PCR process is automated, said Gilad Almogy, Ultima's CEO, but a person must manually load a tube with the library into the sequencer.
Not everyone was as averse to emulsion PCR as Pellegrino da silva, but she wasn't alone. She noted that her experience was colored by the fact that during her postdoc, her lab bought about a dozen SOLiD sequencers, which ultimately did not yield any data for use in a publication. "It was a disaster," she said.
Almogy suggested that the firm has improved upon earlier attempts at emulsion PCR by automating the process and "[freeing it] of any nasty chemicals."
"Everybody asks us about it once," he said.
Ultima's problems with homopolymers, specifically runs of the same 11 or more bases, an issue also seen in previous flow-based chemistries, were also of concern. Attendees passed around phones to show an extensive Twitter thread from Lior Pachter, a computational biologist at Caltech who went hunting for data generated on Ultima's system over the past week.
"The technology is terrible at and around such homopolymers (and even shorter ones)," he wrote. "The indel situation is particularly bad."
"There's no paired-end reads and it seems limited in its ability for structural variation identification," said Bruce Kingham, a core lab director at the University of Delaware, right after lunch on Wednesday. "It's meant to compete with NovaSeq, but there was nothing [in the sponsored lunchtime talk] about the core of the technology. That's a bit concerning and suggests it's still a work in progress."
Chris Lytle added that his small core lab at the Dartmouth Cancer Center might not be the target market for Ultima but suggested that large production-scale labs could use it. "This is a 23andMe machine," he said, alluding to the consumer genomics firm, which currently does not use sequencing but has more than 12 million customers worldwide.
Kingham later said he visited the Ultima suite and learned more about the technology and company. "They seem well positioned to satisfy the needs of the segment they're trying to address," he noted.
The Broad Genomics Platform has had two Ultima systems since 2019, lab director Stacey Gabriel said during a sponsored talk. Per-base accuracy was better than Q20 for 95 percent of bases and better than Q30 for about 85 percent, she reported.
The current run time for the platform at Broad is a little more than 18 hours for 464-flow runs, generating between 8 billion and 10 billion reads per run.
By the end of summer, the Broad plans to use Ultima for profiling applications such as single-cell sequencing and perturbation screens, as well as for cancer cell line characterization and a population sequencing cohort. Gabriel noted that she's looking for improvements in chemistry, base calling, and variant detection to "narrow current gaps in quality."
This week, Ultima also announced partnerships with Nvidia to embed that firm's graphics processing units into its UG100 sequencing instruments and with Olink to add NGS-based proteomics as an application.
As with Ultima, the Broad Institute was also one of the first labs to get access to Element's Aviti sequencer. The Broad test-drove Aviti for its "walk-up sequencing service," which offers fast turnaround sequencing for researchers on various sequencing platforms, according to Sheila Dodge, general manager of the Broad Genomics Platform.
The lab evaluated the platform for a variety of metrics using whole-genome sequencing, including workflow scalability, stability and failure rate, data quality, and human variant calling quality.
During three weeks of beta testing, the lab observed accuracy of Q30 for more than 90 percent of bases across all runs, with an average Q score of 39. The average yield was between 130 Gb and 150 Gb per flow cell and all nine runs were successful. A Broad Institute spokesperson later said that the presentation had a typo and the lab saw out put of 130 Gb to 150 Gb per flow cell lane. "There are two lanes per flow cell and we indeed achieved up to 300 Gb per flow cell," they said, in line with Element's official flow cell specifications.
Dodge said there are a few features she likes about Element's platform, for instance its flexibility, enabled by two independently run flow cells, multiple sequencing kits, as well as different viewpoints and sequencing modes. Additionally, she said the cost is "game changing," although she did not disclose the specific cost per sample.
Moreover, she said the data quality is "essentially equivalent to what we see on our own machines." Specifically, her lab compared PCR-free whole-genome sequencing data quality for GIAB HG001 and HG002 samples between Illumina NovaSeq and Aviti. The results showed highly concordant data in terms of SNPs and indels between the two platforms.
In an analysis of variant calling in whole-genome sequencing using Google's DeepVariant, Element's performance regarding the number of false candidates was equivalent to Illumina's at 30X coverage, even when using an artificial intelligence model trained for Illumina data, Google's Andrew Carroll said in a talk sponsored by Element. Using a model trained on Element data provided a small boost, he added, and at 15X coverage, Element had about 30 percent fewer false candidates than Illumina.
"There is some cleanness or noise reduction," Carroll said. "There's something in the underlying data that does appear to have an advantage [over Illumina]." Somatic variant detection, cell-free DNA sequencing, and low-coverage sequencing applications may all be able to take advantage of that, he suggested.
Encouraged by the results, Dodge said the Broad has purchased three Aviti platforms for its walk-up service with one already on-site. She also pointed out that the lower throughput of Aviti compared with Illumina's NovaSeq could promise more leeway for the lab to achieve faster sequencing turnaround times.
DNA input requirements for converting Illumina libraries are still high, she noted, adding that Element also needs to enable more cloud vendors.
Some were not immediately sold on the Element platform, though. "It seems a little complex," University of Delaware's Kingham initially said of Element's chemistry. However, after talking with the company, he said he realized "it's not necessarily more complex than the chemistry I'm currently working on," noting that he believes it has "led to significant improvements in data quality."
So far, Element has shipped four instruments, according to Chief Technology Officer and Cofounder Mike Previte, who added that the company has run 11,000 sequencing runs in house on 28 instruments. Element added that New England Biolabs has also purchased one.
The firm revealed plans to offer 2x75 bp kits in Q4 of this year, with a price tag of $1,080 per kit, or about $1 to $1.34 per million reads, for single-cell users. This is a common read length in single-cell sequencing, according to Irina Matetovici, a sequencing technology specialist at Belgium's VIB Tech Watch team. "Usually, people fit it in 100 bp for 10x Genomics [single-cell] libraries," she said. "For other methods with a lot of barcodes, read lengths may need to be longer." Element's kit could represent as much as a 50 percent discount to what it costs on a comparable Illumina machine, she said.
Element said the Broad's average flow cell output was reduced by several runs of those 2x75 bp kits, noting that their official specification is 240 Gb per flow cell. The Broad did not immediately responde to a request for comment.
Kingham said he has been most impressed with Singular Genomics' platform. "Element has the flexibility of two flow cells. That's good, but Singular's flexibility would be more advantageous to me. It's a big, big deal to have four independently addressable lanes."
Dartmouth's Lytle agreed, saying it would eliminate the complexity of having to mix barcodes and samples from different studies together on a larger run on, say, a NovaSeq. "I like their chemistry a lot more," he added.
In a talk Wednesday night, PacBio CSO Jonas Korlach provided the first specs for the firm's sequencing-by-binding short-read technology, which it plans to offer in a "mid- to high-throughput" instrument. The firm has 12 instruments running "around the clock" in house as it makes improvements to the method, which features separate base interrogation and incorporation steps. Korlach said that 90 percent of bases are at or above Q40, or a 1 in 10,000 per-base error rate. Single nucleotide variant calling had 99.7 percent precision (the number of correct positive results divided by the number of all positive results) and 99.6 percent recall, while indel precision was 99.4 percent and had 99.9 percent recall. Korlach noted that he expects improvements to those specs through variant caller training.
PacBio has logged over 1,500 runs on the new platform, he said, and has improved its run success rate to 95 percent, up from 70 percent in February.
Beyond academic core facilities, other kinds of labs are welcoming the new sequencing platforms.
"Having this many sequencing options on the market and the cost constantly going down is always good for users like us," said Sangmoon Lee, chief technology officer of Genome Insight, a whole-genome sequencing data curation startup based in San Diego.
However, Lee said, his company wants to see Ultima or Element data on cancer samples paired with normal tissue for detecting somatic mutations.
Ultima offers "a promising technology that can potentially [improve] the clinical utilization of whole-genome sequencing," said Pengfei Liu, associate clinical director for molecular and NGS at Baylor College of Medicine, which is also an early-access customer of Ultima.
In a separate talk at AGBT, Liu highlighted preliminary data of a comparison between Ultima and Illumina whole-genome sequencing using GIAB samples.
"Our first experience looking at the unfiltered, unprocessed raw [Ultima] data is that for SNPs, we're doing good," Liu said. For clinical diagnosis of rare disorders from germline data, the homopolymer and indel issues do not have a large impact because they tend to be regions that have benign polymorphisms, which are less likely to have a clinical impact, he added.
Overall, he said Ultima delivered "very positive results" compared to Illumina sequencing and was able to successfully detect diagnostic variants. What's more, he said, the low cost enabled his team to perform deep RNA sequencing, as well, which could improve diagnoses.