NEW YORK (GenomeWeb) — At the Clinical Virology Symposium in Daytona Beach, Florida last week there was a buzz among attendees about point-of-care nucleic acid testing and multiplex syndromic panels. But, amid all the discussion of PCR-based molecular diagnostics, a presentation on unbiased metagenomic sequencing stood out.
Charles Chiu, a professor of laboratory medicine at the University of California, San Francisco, told an audience of virologists and clinical lab directors that a method developed for the Illumina MiSeq may ultimately be ported to Nanopore's MinIon, teasing the potential for real-time sequencing-based pathogen identification in minutes.
Unbiased, agnostic sequencing marks a true "paradigm shift" for clinical labs, Chiu told GenomeWeb in a follow-up interview this week.
Clinicians currently formulate differential diagnoses, then order targeted tests, or, occasionally, small multiplex PCR panels.
With metagenomic sequencing — which aligns randomly sheared short DNA sequences to reference taxonomies — clinicians would no longer be targeting individual pathogens or small menus of infectious viruses, bacterium, fungi, or parasites.
"We're using a single test that can encompass all known pathogens that cause disease," Chiu explained.
As previously reported by GenomeWeb, the group intends to launch a workflow for laboratory developed tests combining agnostic metagenomic sequencing on Illumina's MiSeq with an informatics pipeline called sequence-based ultra-rapid pathogen identification, or SURPI, which was described last year.
That workflow was used to make a neuroleptospirosis diagnosis described last year in the New England Journal of Medicine. It can be run on a wide range of sample types in a CLIA-certified clinical lab, and the group is honing it to use as many automated steps as possible, Chiu said. The current turnaround time is 12 hours to two days, but the group hopes to get it down to eight hours so it can fit in a single clinical lab shift.
Chiu also noted, for viral sequencing at least, the percentage of sequencing reads appears to be proportional to virus concentration, so the technique could potentially report viral load.
The initial LDT will likely be for more narrow clinical indications, such as culture confirmation or viral meningitis/encephalitis, and may be available in three to six months.
"We want it to remain focused, because it makes the task of validation much more straightforward," Chiu explained.
Because there isn't any precedent for a metagenomic assay used as an LDT, narrow indications will also help the group to see how such a test will perform in real life situations, he noted.
Furthermore, Chiu described a parallel method of transcriptome profiling on MiSeq that could be used to report signatures of host response to infection in the same samples.
"We are actually implementing that right now, and plan to develop that initially as a separate LDT, but the long-term goal would be to combine it into a single test," Chiu said, noting that this could be complementary to the direct detection data.
A publication currently under review from Chiu's lab will show that, for certain infections, pathogens can be identified "down to the species level" solely on the basis of the host response, he said. The group is also working on novel sample prep methods to improve throughput and yield in the workflow.
But metagenomic diagnostics are also uncharted territory for regulatory agencies, such as the US Food and Drug Administration, Chiu noted.
"It's obviously going to be impossible to validate a test this broad for every single pathogen. … There may have to be new models of regulatory validation developed so that these tests can actually become FDA-cleared and make it into the market," Chiu said, adding that he continues to discuss these issues with the FDA.
However, he also suggested that perhaps validating an unbiased approach — which uses no primers or probes — may not be so challenging after all, if it can be simply seen as a single standard protocol.
In his CVS presentation, Chiu showed preliminary data from Ebola patient samples, and thus an Emergency Use Authorization may be pertinent if metagenomic sequencing is used to detect Ebola virus, he said.
"FDA approval, through a [premarket approval] or EUA, depending on the application, is going to take time, but at least we are on the path towards that," Chiu said.
New MinIon data
In his CVS presentation, Chiu also showed preliminary data on metagenomics-based diagnostics using Oxford Nanopore's MinIon.
"We've taken the same metagenomic protocol that we developed for the MiSeq and ported it over to the nanopore … we're hoping to get a sample to answer turnaround time from 12 to 24 hours down to under six hours," Chiu said.
He also presented MinIon data showing real-time sequencing of patient samples with known infections.
Specifically, he said his group had developed an algorithm to classify reads "on the fly" and showed a sped-up movie of the method identifying Ebola virus in five minutes and 15 seconds.
Chikungunya virus was detected from a high titer sample in about seven minutes.
"We were not only able to identify Chikungunya virus on the nanopore, but … by taking advantage of the redundancy in the coverage we're able to assemble most of the genome with 98 percent accuracy," Chiu said during the presentation. This method can also identify the strain and even enable SNP-level analysis of the genome.
However, the real-time method also misidentified snippets of human DNA as belonging to gorilla, mouse, rat, cow, and crow genomes.
This is due in part to high error rates on the Nanopore. "They are coming down, but they are still in the range of 10 to 30 percent," Chui said.
"Because of that, if you're using single reads and aligning them against a reference database like GenBankNT, you will get genuine human reads that, on the basis of alignment, also have homology to various eukaryotes," he explained.
For infectious disease diagnostics this could be problematic if the pathogen is a worm or other eukaryote. For viruses, however, it may not matter.
"Their sequence space is actually quite specific to viruses, and, given the diversity of viruses even within individual families, genera, or even species … we've been able to show you can make unambiguous identification even on the basis of a single error-prone read," Chiu said.
The MinIon also has the disadvantage of relatively low throughput. Chiu's group has been generating about 100,000 reads per run, of which only about 10 percent tend to be good. "Ten thousand reads … is really challenging for groups doing metagenomics," he said.
During the talk, he also noted his group has recently started to do dilution experiments to determine limits of detection, and is currently able to achieve an LOD of 104 target copies per ml. "I don't think we can get to … PCR-level sensitivity with this current throughput," he said.
The MinIon is essentially a single-flow cell, and although Oxford Nanopore hasn't yet provided access to a prototype of its PromethIon, to Chiu's knowledge, clustering many of these flow cells together may help solve the throughput problems.
"With the nanopore, we're still exploring research applications — eventually the goal would be to develop this as an LDT, but we're not close to that yet," he said.
In addition to potential improvements in the Oxford Nanopore technology, which is "getting better and better," Chiu also noted that his group is developing algorithms for improved taxonomic classification. "I think that there are ways that we can deal with this issue, especially if you want to detect eukaryotic pathogens such as plasmodium," he said.
The timeline
Initially, viral hemorrhagic fever or viral febrile illness might be well-suited applications for a point-of-care diagnostic based on the Oxford Nanopore platform. To get a sense of the paradigm shift, consider that a syndromic multiplex PCR panel for fever would have to choose targets carefully, be vigilant in monitoring mutating pathogen genomes, and would likely be challenging to validate for rare organisms.
Different elements of the current metagenomic method have intellectual property protection via the UCSF tech transfer office, Chiu said, and there is potential to license aspects of it, such as the interpretation software for SURPI. The exact composition of the host expression panels also comprises IP that could be licensed or commercialized, he said.
But Chiu wishes to make the method accessible. "I really want to see this disseminated as rapidly as possible, so I'm hoping that [IP] doesn't impair the transfer of the knowledge and information," he said.
The potential to save patients' lives and provide diagnosis in tough situations — such as the up to 80 percent of encephalitis cases in which a pathogen is never detected — motivates Chiu. Dissemination may happen through industrial collaboration, he said, noting his group has received interest from "various companies" about the software, procedures, and panels.
"Hopefully, this might be a springboard by which companies can … bring these assays into a clinical setting or into public health. This is why I'm very eager to work with them and help to develop this and get it out there."
After the talk, some in the audience at CVS seemed to think that NGS-based pathogen ID is more than five years away from clinical use, but Chiu begs to differ.
"I know that some people are skeptical. A metagenomic assay being FDA approved may certainly be a few years off, but … we've already show in principle that it can be used to give a clinically relevant diagnosis," he said.
"I really think that this is not a matter of five years, but rather within a year or two you will start seeing, from not only my lab but other labs as well, LDTs that are based on metagenomic sequencing."