Using AI to fight Coronavirus
As scientists make strides in finding answers about COVID-19, artificial intelligence has aided one Michigan State University researcher and his team in finding answers about the new omicron variant. The MSU researchers report omicron and other variants are evolving increased infectivity and antibody resistance according to an artificial intelligence model. Therefore, new vaccines and antibody therapies are needed, the researchers say.
Understanding how SARS-CoV-2 evolves is essential to predicting vaccine breakthrough and designing mutation-proof vaccines and monoclonal antibody treatments. In a recent study in the journal American Chemical Society Infectious Diseases, Guowei Wei, MSU Foundation Professor in the Department of Mathematics in the College of Natural Science, and colleagues, analyzed almost 1.5 million SARS-CoV-2 genome sequences taken from people with COVID-19.
They identified 683 unique mutations in the region of the SARS-CoV-2 spike protein that attaches to the human ACE2 receptor on the surface of human cells for virus cell entry, which initiates the infection. Then, they used an AI model to predict how these mutations affect the binding strength of spike protein and ACE2 as well as spike protein and 130 antibodies that created from prior infection or vaccination to prevent future viral infection. Several antibodies authorized by the FDA as COVID-19 therapies were also included in the study.
The team found that mutations to strengthen infectivity are the driving force for viral evolution—a process in which the most competitive variant is selected for dominancy—whereas in highly vaccinated populations, mutations that allow the virus to escape vaccines become dominant. The researchers also predicted that certain combinations of mutations have a high likelihood of massive spread.
“With this AI model we can predict how infectious each variant is, how often vaccinated individuals become infected when exposed to the virus, and how well vaccines protect against new variants without using extra experimental data,” Wei said.
In another study in the Journal of Chemical Information and Modeling, Wei and colleagues took a deep dive into the omicron variant’s level of contagiousness, vaccine breakthrough and antibody resistance. They used their AI model to analyze how the variant’s unusually high number of mutations – 32 – on the spike protein affect receptor-binding domain, or RBD, which directly binds to ACE2 and antibodies. RBD is a key part of a virus located on its spike protein that allows it to dock to body receptors to gain entry into cells and lead to infection.
Their results indicated that omicron is over 10 times more infectious than the original coronavirus and 2.8 times more infectious than the delta variant. In addition, omicron was 14 times more likely than delta to escape current vaccines, and it is predicted to compromise the efficacy of several FDA approved antibody therapies.
“Many of these predictions have been verified by emerging experimental results, stressing the importance of developing a new generation of vaccines and monoclonal antibodies that won’t be easily affected by viral mutations,” Wei said.
Banner image: A team of Spartan researchers, led by MSU Foundation Professor Guowei Wei, report that omicron and other variants are evolving increased infectivity and antibody resistance according to an artificial intelligence model. Therefore, new vaccines and antibody therapies are needed.
