ORLANDO, Florida – Ultima Genomics has closed the early-access program for its UG 100 high-throughput sequencer and is ready for a full commercial rollout, CEO and Cofounder Gilad Almogy told attendees of the Advances in Genome Biology and Technology meeting on Tuesday.
How big that rollout will be is the ultimate question. With the ability to process up to 20,000 whole human genomes per year, the instrument is likely to appeal mostly to the largest sequencing labs around the world. Those who have the samples to feed the machine will enjoy the pricing, previously disclosed, of $1 per gigabase or approximately $100 for a human genome.
Several high-profile customers have already gotten final upgrades to their early access units, including the Broad Institute Genomics Platform (which has two), the New York Genome Center, Exact Sciences, Quest Diagnostics, South Korea's Macrogen, Genome Insight, and pharma firm Regeneron — also an investor in Ultima.
In scientific talks following Almogy's unveiling of the final launch specs, customers praised the improvements seen in accuracy, ease of use, and other key factors.
Broad Genomics Platform Director Stacey Gabriel described how her facility is now in the "early stages of production sequencing" with two UG 100s. Her lab was the first to engage with Ultima, back in 2018. She highlighted improved stability, productivity, and quality of the runs. The overall workflow has gone from four days down to two; reads passing filter have increased over time from 6 billion per run to 8 billion. Variant calling has also improved, especially for indels, which now have an average F1 score, a measure of both precision and recall, of 99.28 percent for most of the genome.
Single-cell sequencing expert Rahul Satija, of the New York Genome Center and New York University, presented data from CRISPR-based perturbation screens. "We get exactly the same biological insights" with Ultima as with Illumina sequencing, he said. "We're fully comfortable switching our perturb-seq experiments entirely."
Will Salerno of the Regeneron Genetics Center showed how Ultima has allowed the company to quickly create a resource for data imputation in their exome sequencing cohorts. The platform is literally on its way up in the world. "Up until now, it has been in the basement," he said. "Now, it’s being pulled upstairs into the room with glass walls that everyone sees when they walk by."
This week's AGBT meeting, where Ultima is the top sponsor, marks a year and a half since it came out of stealth just ahead of the 2022 meeting, held in June of that year. The firm's main message has remained the same: that it will provide competition for Illumina in the high-throughput, short-read sequencing market, starting at a price point of $1 per Gb. At that time, that was almost six times cheaper than the Illumina NovaSeq 6000 with an S4 flow cell; however, Illumina's NovaSeq X with the 25B flow cell now offers sequencing with a list price as low as $2 per Gb.
Ultima also raised eyebrows by revealing that it would use a flow-based chemistry, an approach that has been tried before and struggled with homopolymers — stretches of the same base all in a row — as well as emulsion PCR for amplification that takes place on a separate instrument.
The company hasn't changed its approach to library preparation, but Almogy said he's confident the firm has solved the problems with homopolymers and the problems they cause in indel accuracy. "The question was, ‘Will we ever get to 99 percent [indel F1 accuracy]?'" he said. "We've blown past that."
SNV F1 has always been well above 99 percent and at launch was claimed at 99.8 percent, while indel F1 has reached 99.4 percent, he said.
Gabriel presented data generated at the Broad showing that for more than 99 percent of the Genome In A Bottle High Confidence Region (GIAB HCR), indel F1 averaged 99.3 percent over seven samples. For the entire GIAB HCR, average indel F1 was 94.9 percent.
In an email, an Ultima spokesperson said the adjusted HRC "excludes 13-mers and above and almost the entirety of the exclusions sit outside the exome."
The improvements are due to better handling of homopolymers between 8 and 12 bases long. Eight- and nine-mers are 99 percent accurate while 12-mers are still 96 percent accurate. "For any practical purposes, for any coding homopolymers, we've closed the gap," Almogy said. Ultima improved its performance by tweaking the chemistry and with more advanced use of artificial intelligence, he said.
Along with the technical specs, Almogy touted the UG 100 workflow and flexibility. "We learned our customers really like automation," Almogy told GenomeWeb ahead of the lunchtime session. The UG 100 can sequence continuously, with up to six of its wafers loaded at a time and two wafers sequenced simultaneously. With run times of about 12 to 14 hours, each machine can produce up to 50 Tb of data per week.
At $2,000 each, the wafers provide 6 billion to 8 billion reads, or approximately $.30 per million reads.
Almogy compared the price of Ultima's wafers with Illumina's NovaSeq X flow cells, noting that each wafer costs less than the smallest X flow cell, while providing a lower cost per Gb than the biggest.
In addition to meeting its promises to the field, Almogy said Ultima is also adding something people may not have expected: a workflow that provides raw error rates for base substitutions of 1 in 1 million, or Q60 accuracy. "This is without over-sequencing, without unique molecular indices (UMIs)," he said. "It's our standard sequencing run but with a slightly different library prep kit and different analysis."
Dubbed PPMseq (Paired plus minus sequencing), it is similar to existing duplex sequencing methods, except it captures both strands of DNA onto the same emulsion PCR bead, explained Mirna Jarosz, VP of product strategy at Ultima. "When you denature them, we keep them together in the emulsion," she said. "There's no molecular barcoding, no read collapsing. It's coming out in the raw read."
"You start to see what's sequencing quality and what's sample quality," she said.
Dan Landau, an oncologist at Weill Cornell Medicine, presented data from his lab's use of PPMseq for duplex sequencing of the whole genome. "This has rarely been attempted," he said. "Duplex sequencing at the genome level was considered cost prohibitive."
With other duplex sequencing methods, the complementary molecules need to be captured together by chance, resulting in only 3 to 4 percent of the library yielding duplex data. But PPMseq adds barcodes to double stranded DNA on the sequencer, so the two strands are analyzed together. "Now, we don't have to have the two strands find each other," he said. "Instead of 3X coverage, we get 50X duplex coverage from a 100X genome."
Chris Mason, a sequencing expert at Weill Cornell, said the UG 100 is "one of the biggest steps forward in genomics in the past five to ten years." His lab is using access to the NYGC's instrument for whole-genome sequencing, metagenomics, and epigenetics. And "some astronaut genomics, of course," he said, referring to his lab's interest in the effects of spaceflight on human biology.
"Now we want to see deployment of the machines [with] reliability confirmed beyond two to five machines and get up to 20, to 50, to 100 machines," Mason added.