Artificial Intelligence Nanopore for Detecting Different Coronaviruses

28-07-2021 |   |  By Liam Critchey

The current pandemic, COVID-19, brought about by the coronavirus strain SARS-CoV-2 has been ongoing around the world for over 18 months now. When the pandemic took the world by surprise, there was a limit on the number and types of tests available. However, as the months have gone on, the standard PCR test has reduced from week-long analysis times to following day results. Other tests, from lateral flow to adapted microfluidic systems, and point of care biosensors, have all emerged (although only certain ones have been produced on large enough scales to be usable).

Despite the pandemic being unexpected, it is not the first coronavirus to impact the world. However, other coronaviruses such as the SARS, MERS and original human strain HCoV-229E have created localised epidemics worldwide. SARS-CoV-2 is the only strain that has facilitated the development of new testing methods, and these developments are still ongoing.

Current Gold Standard Testing

The SARS-CoV-2 strain is the seventh strain of novel coronaviruses that we have seen. It has required a multi-faceted approach to try and prevent further outbreaks during this pandemic and prepare for any similar coronavirus strains that could affect the world in the future. A future society that would hopefully be fully equipt after all the R&D that has taken place during this pandemic). While the global vaccine drive is a significant factor in stopping the pandemic and preventing people from getting seriously ill, continuous and ongoing testing is still a meaningful way to prevent the spread of coronavirus to other people.

Polymerase chain reaction (PCR) testing has been the gold standard throughout the pandemic and remains so to this day. It remains so because some of the limitations of week-long analysis times early in the pandemic have shortened. We indeed need other tests, such as the lateral flow tests for rapid testing, but PCR has become the gold standard for more accurate readings.

PCR is a test where the RNA needs to be extracted and amplified and requires much active handling from the analyser. So, there is always a risk of exposing the handler(s) to the coronavirus. Active handling is one of the main negatives of the test, despite the good results it produces. The scientific community is interested in developing tests that require less user handling, so the risk of coronavirus exposure reduces. Over the last year, one option has involved modifying closed microfluidic devices that can detect SARS-CoV-2. However, a different way has emerged recently: to combine nanopore sensors with machine learning algorithms.

Nanopore Sensors with AI

Nanopore sensors can utilise nano-sized pores ranging from several nanometres to several hundred nanometres. Several different nanopore sensors are already out there that can detect different biomolecules such as DNA, viruses, and bacteria. Viruses that get detected by nanopore sensors get transported from the cis side to the trans (chemically speaking) of the nanopore by an electrophoretic force.

When this happens, the ionic current across the nanopore decreases. Artificial intelligence can be used herein the analysis because the ionic current vs time waveform—which can be obtained from the nanopore—contains information about the volume, structure, and surface charge of the virus being analysed. By analysing this data with AI algorithms, a single virus can be identified with high accuracy without extracting the genome.

While there is much promise for this type of analysis, there are still some challenges with this approach, especially surrounding virus learning data for the AI algorithms and obtaining precise detection with an increased reproducibility. Additionally the ionic current characteristics of the nanopore rely on the electrical characteristics of the nanopore, the electrical characteristics of the measuring device, and the fluidic device that transports the patient’s sample to the nanopore. So many of these sensing systems have been missing a dedicated measuring device and flow channel suitable for the nanopores. However, a research team has now built a complete AI-assisted measurement platform to detect viruses accurately.

Creating a Complete Platform

This complete coronavirus sensing platform comprises semiconductor nanopore devices, portable and high precision current measuring devices and is used in conjunction with machine learning algorithms. The platform does not require RNA extraction in any form, so it offers a safer way of handling coronavirus testing. The detection system detected a range of different coronavirus strains, including HCoV-229E, SARS-CoV, MERS-CoV, and SARS-CoV-2.

In this approach, the training data obtained from cultured viruses was changed to detect PCR-positive and PCR-negative samples. Using the training data in this way enabled the AI-nanopore system to detect both positive and negative samples with high sensitivity and high throughput, turning it into a versatile virus diagnostic platform.

The platform was an ideal solution for determining if a patient is showing symptoms of a coronavirus strain or whether they have a different virus that shows similar symptoms, e.g., influenza. At first glance, influenza A strains show similar flu-like symptoms to SARS-CoV-2, so if the symptoms go undiagnosed, it can lead to greater exposure and risk to both clinicians and the general public. Testing was a problem early on in the COVID-19 pandemic, where there were a higher number of false-negative and false-positive results.

The machine learning element of the diagnostic platform distinguishes between SARS-CoV-2 strains and influenza A (H1N1) viral strains by learning the difference from cultured samples. The platform showed that it could discriminate between the two different viruses. Learning of the cultured versions from patients saliva samples could identify if a patient is infected by either the SARS-CoV-2 or the influenza A virus.

Future Outlook

Given the current pandemic, there was much focus on the efficiency and ability to detect SARS-CoV-2 strains and how testing can distinguish between a patient with coronavirus over a patient with similar symptoms but a different virus. The platform detected SARS-CoV-2 in a saliva sample with a sensitivity of 90% and specificity of 96% in just 5 minutes.

Despite the focus on SARS-CoV-2, the platform was suitable for all the main coronavirus that caused epidemics over the years. If scientists are correct, then SARS-CoV-2 may not be the last coronavirus to cause an outbreak. So, developing tools for the future to accurately detect different coronaviruses (and not just specific strains) will be vital to getting on top of any outbreaks, should they occur again, so that any outbreaks could be contained and dealt with quickly. While many suitable tests can be used already, the more tests and different types of tests we have at our disposal (providing they are commercially feasible for large-scale manufacturing), the better equipped we may be.


Taniguchi M. et al, Combining machine learning and nanopore construction creates an artificial intelligence nanopore for coronavirus detection, Nature Communications, 12, (2021), 3726

Liam Critchley Headshot.jpg

By Liam Critchey

Liam is a science writer who specialises in chemistry and nanotechnology, and reports on the extensive amount of areas which cross-over with these disciplines. As a writer, Liam has worked with companies, media sites and associations around the world and has published over 600 articles to date. Liam is also a member of the advisory board for the National Graphene Association and the Nanotechnology World Association and is a member of the board of Trustees for the charity GlamSci. Before becoming a writer, Liam obtained two masters degrees in Chemistry with Nanotechnology and Chemical Engineering.

Related articles