AI predicts dangerous virus mutations before they happen
The program can be used to develop vaccines and drugs. The study was published in the journal Nature.Researchers from Harvard Medical School and Oxford University have created a system based on artificial intelligence called EVEscape. The program predicts the evolution of a virus by predicting the most likely mutations and new strains of the virus that will appear in the future.
EVEscape consists of two elements: an evolutionary process model that predicts the changes that may occur to the virus, and detailed biological and structural information about the virus. Together, these components allow EVEscape to predict how the virus might evolve.
The researchers tested the model's performance in predicting coronavirus mutations. The scientists loaded only data up to January 2020 into the system, and then watched the AI predict mutations that actually occurred after that date.
EVEscape predicted new versions and their prevalence more accurately than the experimental methods currently in use. The AI did not have to wait for antibodies to the new strain to develop in diseased individuals and become available for testing. Therefore, the program made predictions faster and more efficiently than laboratory testing. The AI also made accurate predictions about other viruses, including HIV and influenza.
The technology was able to sort through the tens of thousands of new SARS-CoV-2 variants emerging each week and identify those most likely to cause problems. The researchers are now applying EVEscape to the virus in real time and publishing an updated forecast every two weeks.
The program could potentially serve to proactively develop vaccines. Based on the predicted mutation, companies could evaluate efficacy and modify existing vaccines in advance.