A new machine learning model developed by researchers at Michigan State University suggests that mutations in the SARS-CoV-2 genome made the virus more infectious. The research results have been published on the ArXiv preprint server.

The model, developed by lead researcher Guovei Wei, a professor in the Department of Mathematics, Biochemistry, and Molecular Biology, analyzed the genotyping of SARS-CoV-2 from more than 20,000 viral genome samples. The researchers also looked at mutations in the spike protein – a protein primarily responsible for spreading infection – and found that five of the six known subtypes of the virus are now more infectious.

As with any virus, many mutations end up being benign, with little or no risk to infected patients. Some mutations even reduce infectivity. But some of them make the virus more contagious.

Wei and his team spent several months studying and analyzing the patterns and sites of mutations, tracking changes from the official sample of the viral genome obtained in January.

Knowing the infectivity of SARS-CoV-2 is vital to prevent COVID-19 and restore the global economy. The key question is what are the implications of these mutations for the transmission, diagnosis, prevention, and treatment of COVID-19.

Guo-Wei Wei, Professor of the Department of Mathematics, Biochemistry and Molecular Biology

Viral infection occurs when a spike protein interacts with a receptor in a human host cell, angiotensin-converting enzyme 2, abbreviated ACE2. Concerning ACE2, scientists are concerned about a concept known as binding affinity, or the strength of the binding interaction between the spike protein and the host receptor in the early stages of infection.

The infectivity of the virus increases if the binding affinity increases. More than 50 mutations have now been found, as well as a binding interface on the receptor-binding domain of spiked proteins – RBD for short, which has 194 possible mutation sites, the researcher explains.

Wei’s Machine Learning Model, an advanced neural network, analyzed over 8,000 records of protein interactions to determine the effect of current known mutations on the binding affinity of the SARS-CoV-2 spike protein. The result, which suggested increased binding affinity in five of the six known subtypes, indicated that infectivity could have increased as a result of mutations.

Concerned about the possibility of further mutation, Wei and his team turned to their model of the future.

It is extremely important to know if future SARS-CoV-2 subtypes will pose an imminent threat to public health. To this end, we performed a systematic screening of all possible 3,686 future mutations at 194 sites of possible mutations along the RBD.

Guo-Wei Wei, Professor of the Department of Mathematics, Biochemistry and Molecular Biology

Wei’s model predicts that multiple residues in the receptor-binding motif – a component of the RBD region – have a high chance of mutating into more infectious COVID-19 strains.

He warns that while the predictions based on artificial intelligence are consistent with the available experimental data, further research is needed to fully understand the impact of mutations on COVID-19 infectivity.

As part of their research, Wei and his team also predict that the new coronavirus spreading around the world is slightly more contagious than the original SARS virus discovered in 2003.

Wei said the results are consistent with the results of another experiment recently published by scientists at the Scripps Research Institute in Florida. This study examined protein spike mutations in vitro and found that the virus mutated in such a way that its infectiousness increased.