NEW YORK ─ University of Oxford researchers are developing a new diagnostic method that leverages single-particle imaging and deep learning to return test results for active SARS-CoV-2 and other respiratory infections within five minutes.
The method, published as a MedRxiv preprint, which has not been peer reviewed, could enable additional rapid testing that would ultimately help clinicians to better differentiate among seasonal respiratory infections in point-of-care settings, Nicolas Shiaelis, the first author of the paper and a researcher at the University of Oxford, said in an interview.
Its enabling technology would give it an edge over antigen and RT-PCR tests currently on the market to detect SARS-CoV-2 active infections, he said, adding the group believes tests with such rapid turnaround times could help mitigate the pandemic, as cases continue to soar worldwide.
"If you can do testing in less than five minutes, you'll have better access to big venues and testing at company sites to help start up the economy again," Shiaelis said.
He believes that the technology holds particular promise for use during flu season, when it is difficult to distinguish whether a person has the flu or SARS-CoV-2.
Current tests for active SARS-CoV-2 infections detect pieces of viral particles ─ specific virus proteins in the case of antigen tests and RNA in the case of RT-PCR. However, the University of Oxford method gets its performance and speed by detecting whole, or intact, particles that don't need to be broken down or segmented, Shiaelis said.
The method eliminates lysis, RNA purification, and PCR amplification steps that are necessary in RT-PCR testing, shortening the time to results. Further, removing those steps eliminates reliance on reagents that are sometimes in short supply and constrain the availability of RT-PCR testing, Shiaelis said.
To detect intact particles, the University of Oxford system connects single-stranded DNA with fluorophores to the virus, enabling the detection of single particles.
The system uses fluorescence imaging to capture the interaction of the combination of single-stranded DNA with virus particles, and machine learning software developed by Shiaelis analyzes the images and confirms whether SARS-CoV-2 or another targeted virus is present.
The system achieves labeling ─ the identification of target particles ─ by introducing calcium chloride. "When viral particles are present in the sample, positively charged calcium cations mediate an interaction between the negatively charged virus membrane and the negatively charged phosphate groups associated with single-stranded DNA," Shiaelis said.
The system can differentiate between different respiratory infections because it detects differences in the surface chemistry, size, and shape of particles that are unique to specific viruses.
The researchers collaborated with clinicians at the John Radcliffe Hospital in Oxford to validate the assay in an undisclosed but limited number of samples that were confirmed as SARS-CoV-2-positive using conventional RT-PCR methods.
Prior to evaluating the test's performance, they used clinical samples to train a neural network to differentiate SARS-CoV-2 from negative clinical samples and other common respiratory pathogens, such as influenza and seasonal coronaviruses.
Based on the results of the study, the system has the potential to achieve high sensitivity and specificity on par with the highest performing RT-PCR tests, Shiaelis said.
"In a typical sample, where there are thousands of viral particles, an extremely high sensitivity and specificity close to 100 percent is possible," Shiaelis said. "Our next step is to gather hundreds of clinical samples to verify that we can achieve those numbers."
The method could enable "testing of a lot more people a lot faster than current tests," he said, and his group expects the assay could eventually be applied at point-of-care settings, including workplace settings, music venues, and airports to establish and safeguard infection-free spaces.
The researchers are planning to soon spin out a company to launch a test and to start product development in early 2021. They believe that a test for an undisclosed number of respiratory infections could be available with US and European regulatory approvals around the middle of next year.
To launch the company and complete the development of the test, they have begun working with external business and financial advisors and with Oxford University Innovation, a subsidiary of the university that manages its technology transfer arm.
Though high performing laboratory-based RT-PCR platforms and their assays are the go-to tests to confirm active SARS-CoV-2 infection, the lengthy time to results provides an opening for high performing rapid tests, Shiaelis said.
However, RT-PCR tests performing SARS-CoV-2 testing at the point of care are faster than lab tests, and the University of Oxford test, if it can be launched as a product, would need to compete with tests such as the Abbott ID Now assay, which is already on the market with a US Food and Drug Administration Emergency Use Authorization and delivers test results in five minutes.
The technology developed by the Oxford team remains a long way from the clinic and whether it can be adopted remains to be seen, Cornelius Clancy, a physician and chief of infectious diseases at the US Department of Veterans Affairs Pittsburgh Healthcare System, said in an interview.
The technology is interesting for "its simplicity and the potential ease of testing for clinicians," said Clancy, who has participated in studies to evaluate new diagnostic tests but is not affiliated with the University of Oxford initiative. "However, as always, the proof of the pudding will be in the eating," he said. "How will a commercial test built upon this technology perform with patients in the real world?"
To be adopted, the Oxford test may "need to not only distinguish a targeted pathogen, such as SARS-CoV-2, from other respiratory viruses and the absence of virus, but also accurately identify multiple pathogens," Clancy said.
Its future success could depend on the breadth of its panel and how well it performs for all targets, he said, and added that he believes the advantage of a five-minute turnaround time versus a 15-minute turnaround time for rapid antigen testing is likely to be less relevant in practice.
"If they get the company going, it will be interesting to see what they are able to do with the technology," Clancy said. "I can see a need for a comprehensive, fast, accurate, and cheap respiratory pathogen panel, [but] can they deliver one with this technology?"
Shiaelis said he believes that getting from the current prototype to a product that tests for multiple respiratory infections is achievable.
The current prototype needs further development and its components need to be integrated to form a production prototype, so the group is seeking an undisclosed amount of funding from private investors to support those development efforts.
"What is left is just an engineering challenge ─ to optimize the assay, design the production prototypes, and prepare them for production," he said. "It's nothing that hasn't been done already."
The specific design of a production instrument is yet to be determined. "The final instrument will work autonomously and could, for example, resemble a vending machine where you provide the sample and get a result within a few minutes," he said. "Now that we have good results in the study, [one of the next steps] is to design prototypes to serve the needs of different test settings."
The price of testing is also to be determined, but the researchers are aiming for a price per test and instrument that is "significantly cheaper than RT-PCR testing," Shiaelis said.
He and Nicole Robb, a fellow researcher and test developer at the University of Oxford, are planning the spinout and identifying a leadership team.
They are looking to collaborate with manufacturers or outsource production and are sketching plans for marketing and sales, but "there is nothing concrete yet," Shiaelis said.