Broad Institute Study Evaluates Ultima Genomics Sequencer for Single-Cell RNA-Seq

NEW YORK – Results from single-cell RNA sequencing libraries run on the new Ultima Genomics UG100 are “very similar” to those produced on Illumina’s NovaSeq, a new study says. However, the platform shows bias against 3′ gene expression libraries.

Led by researchers from the Broad Institute as part of a collaboration with Ultima, a multi-institutional team devised a library preparation method to run single-cell libraries generated with 10x Genomics’ Chromium platform on the Ultima instrument, which produces single-end, rather than paired-end, reads. The method uses a PCR-based kit to convert the 10x libraries, which are designed to run on Illumina sequencers, for use with the Ultima sequencer.

“In general, it looks comparable to what you find with Illumina,” said Joshua Levin, a senior group leader at the Broad Institute and corresponding author of the study. The researchers ran single-cell libraries from a number of experiment types, including gene expression, CRISPR-based perturbation screens (Perturb-seq), and cell hashing methods such as CITE-seq (cellular indexing of transcriptomes and epitopes by sequencing).

On key metrics, such as total number of unique molecular identifiers at a sequencing depth of 200 million reads, number of genes or UMIs per cell, and number of cells identified using 10x’s Cell Ranger analysis software, results from Ultima looked like those from Illumina. The researchers published their results Thursday in Nature Biotechnology.

“There are some minor issues that come up, more so with 3′ libraries,” Levin said. Specifically, Ultima sequencing provided fewer total UMIs at equivalent sequencing depth than Illumina. The study attributed this to where the reads came from relative to the gene boundaries and how Cell Ranger deals with those kinds of reads.

“But given that it would be a lower cost, those issues would not be a problem that would prevent you from using Ultima,” he said, adding that for some applications, such as T- and B-cell receptor sequencing, 5′ approaches are required. Ultima has claimed it can deliver NGS of equivalent accuracy to Illumina at approximately $1 per Gb, compared to approximately $6 to $7 per Gb, list price, for Illumina’s highest throughput NovaSeq instrument. “You can change the scale of your screen,” he said.

The paper is an important first step toward making single-cell research more accessible, said Holger Heyn, a team leader at the Spanish National Center for Genomic Analysis (CNAG), who was not involved in the study. Single-cell studies are sequencing-intensive and the cost of NGS has limited how big studies can be. “I think it’s where we have to go, to do more, for cheaper,” he said. “This is how it has to start.”

The study was born out of a high-level collaboration between Ultima and the Broad Institute. Ultima provided the sequencing, Levin said, noting that he was approached by Broad Genomics Platform Director Stacey Gabriel.

At the Advances in Genome Biology and Technology meeting in June, Gabriel presented data from the Ultima platform, noting that the Broad was planning to use it for single-cell research.

Revealed earlier this year after years of secrecy, Ultima’s UG100 sequencer offers single-end reads up to 300 bp in length and 3,000 Gb sequencing output per run.

With its focus on throughput, Ultima aims to compete with Illumina for studies requiring tons of sequencing. But the new study did not compare Ultima data with any data generated on a NovaSeq. New 10x libraries were run on a NextSeq 500, and the Perturb-seq experiment used HiSeq X data that predated the beginning of the study. “In our work, we generally consider all the Illumina sequencers as relatively equivalent,” Levin said. “We may have chosen the NextSeq based on how many reads we needed for a given experiment. There may be other studies that have explored the nuances of the differences among Illumina sequencers, but I’m not fully aware of them.”

Heyn said he appreciated the study, noting that the researchers showed data relevant to “probably 99 percent” of single-cell genomics studies.

The study began by comparing 3′ and 5′ gene expression libraries from peripheral blood mononuclear cells on the respective sequencing platforms. “We obtained similar overall performance,” the researchers wrote, noting that sequencing depths were higher on Ultima. They also tried out single-end Illumina sequencing, which similarly showed “very high agreement.”

The Ultima sequencer did have biases, especially with 3′ gene expression libraries. “One key explanation for the minor differences we observed is the position of reads relative to annotated gene boundaries, as a consequence of Ultima single-end reads being closer to gene ends,” the authors wrote.

There was also a “modest bias … towards genes with higher GC content having higher expression in Illumina and the longest genes having higher expression in Ultima 3′ libraries.”

“Of the 166 genes with differences in expression for 3′ PBMC between the two sequencing platforms, most (130 genes, 78.3 percent) differed in the fraction of reads that were assigned by Cell Ranger to the gene out of all the reads mapped to that gene region,” the authors noted, suggesting that this is related to how single- and paired-end reads map to different locations relative to the transcript. Ultima reads map closer to the 5′ and 3′ end than Illumina reads do. “Because Cell Ranger excludes reads that do not fully map within annotated gene boundaries, more Ultima reads are excluded from analysis as they are closer to gene ends,” they wrote. “This difference in location can also lead to more multimapping or ambiguous reads.”

“There are also some minor quality issues with 3’ sequencing with Ultima that could not be overcome by more sequencing,” Levin said. “These could potentially be improved by making some modification to the 10x construct.” 

These differences “might be a limitation for some studies,” Heyn said. “Historically 5′ libraries were worse, in terms of gene expression profile. But assays are getting better, and now they’re much more comparable.” Levin suggested that 5′ and 3′ gene expression assays are now “interchangeable” for most cases.

And 5′ approaches are necessary for some multi-modal single-cells assays, such as those that add immune cell receptor profiling. For single-cell T- and B-cell receptor assays, 5′ libraries are “the only possible way,” Levin said, as the highly variable regions of interest are in the 5′ region. He noted that the 10x single-cell immune profiling assays would not work with Ultima’s current technology, although longer reads or different library construction could enable that.

The study did not look at single-cell ATAC-seq, a popular assay for epigenetics. Levin suggested that the single-end read limits Ultima to finding only the shortest peaks in open chromatin. “You can still get some information, but not all the information from a typical ATAC-seq experiment,” he said.

“The UG100 is a generic sequencing platform and is capable of accommodating most existing sequencing libraries, including libraries prepared for other platforms via a simple conversion process,” Ultima CEO and Cofounder Gilad Almogy said in an email. “Specifically, we’ve successfully tested sequencing of 10x single cell ATAC-seq and CITE-seq libraries, as well as other single-cell labeling approaches.”

Heyn, who is chair of the technology working group of the Human Cell Atlas project, said the preprint version of the study sparked discussions about how to implement Ultima sequencing into its works. But there are still more studies he would like to see done, starting with a look at more different kinds of tissue and sample types.

He’d also like to see larger studies focusing on immune profiling. “Those studies will be very sensitive to error rates,” he said, due to the highly variable nature of T- and B-cell receptor genes. Errors would artificially increase the variability in those samples, he said. Moreover, he would like to see a benchmarking study that not only compares sequencing accuracy but also investigates whether the sequencing tech results in biases in the interpretation of those data.

Heyn has not yet been in contact with Ultima, but said he wants to invite them to collaborate. “Ideally, we’d run a study as a group,” he said. He’s also interested in testing the platform on its own. “I want to see how it performs in the real world,” he said.

Read more here: Source link