Raman spectroscopy and machine learning method improve detection of SARS-CoV-2


The novel coronavirus, or SARS-CoV-2, which causes the highly contagious COVID-19, has infected millions of people around the world. The global spread of this deadly pandemic has triggered extensive infection control research. However, controlling the spread of COVID-19 is difficult for many reasons.

Some patients present with a variety of non-specific symptoms ranging from headaches to coughing. However, many COVID-19 patients remain symptom-free even after being infected, but may still have the potential to infect others. This makes initial triage and diagnosis difficult. And while reverse transcription polymerase chain reaction (RT-PCR) techniques are currently the gold standard, they have some limitations.

RT-PCR involves transporting samples to a clinical laboratory for testing, which poses logistical challenges. It also requires the use of reagents, which could be scarce and less effective when the virus mutates. Additionally, RT-PCR tests can be time-consuming and less sensitive in asymptomatic people, making them impractical for widespread rapid screening.

Thus, biomedical researchers are trying to design new methods for better detection of COVID-19 infections in healthcare settings, without the need to send samples for testing. Recently, researchers in Canada developed such a technique using saliva samples. Unlike nasopharyngeal swabs, saliva sampling is safer and non-invasive. In their article published in the Journal of Biomedical Opticsthey describe a novel reagentless detection technique based on machine learning (ML) and laser-based Raman spectroscopy.

Raman spectroscopy is commonly used by researchers to determine the molecular composition of samples. Simply put, molecules scatter incident photons (particles of light) in a unique way that depends on the underlying chemical structures and bonds. Researchers can detect and identify molecules based on their characteristic “fingerprint” or Raman spectrum, which is obtained by shining light on samples and measuring the scattered light.

COVID-19 can cause chemical changes in the composition of saliva. Based on this knowledge, the research team analyzed 33 clinically matched COVID-19 positive samples with a subset of a total of 513 COVID-19 negative saliva samples collected at the COVID-19 testing clinic. 19 from Pointe-Saint-Charles in Quebec, Canada. The Raman spectra they obtained were then trained on multi-instance training models, instead of conventional models.

Lead author Frédéric Leblond, based at Polytechnique Montreal, the University of Montreal Hospital Center Research Center and the Montreal Cancer Institute, Canada, puts it more simply: “Our machine learning method uses the information from each Raman spectrum, it does not use average data and therefore can incorporate more information from the saliva samples to give a very accurate result.

The results of this method indicate an accuracy of around 80%, and the researchers found that accounting for sex at birth was important in achieving this accuracy. Although the composition of saliva is affected by the time of day as well as the age of the test subject and other underlying health conditions, this technique may still prove to be an excellent candidate for detection. of COVID-19 in the real world.

Katherine Ember, postdoctoral researcher at Polytechnique Montreal, Canada, and first author of the study, summarizes: “Our label-free approach overcomes many limitations of RT-PCR testing. We are working to commercialize this as a faster, more robust and system at low cost, with potentially higher accuracy.This could be easily integrated into current viral detection workflows, adapted to novel viruses and bacterial infections, as well as accounting for confounding variables through novel approaches to machine learning. In parallel, we are working to reduce additional testing time by using nanostructured metal surfaces to contain the saliva sample.”

These results may facilitate better detection of COVID-19 in addition to paving the way for new tools for other infectious diseases.

Reference: Ember K, Daoust F, Mahfoud M, et al. Saliva-based detection of COVID-19 infection in a real environment using reagentless Raman spectroscopy and machine learning. J. Biomed. Opt. 2022;27(2):025002. do I: 10.1117/1.JBO.27.2.025002

This article was republished from the following materials. Note: Material may have been edited for length and content. For more information, please contact the quoted source.


Comments are closed.