The potential for new waves of illness caused by specific SARS-CoV-2 variants can be forecasted through AI technology. Current models used to predict the dynamics of widespread transmission are unable to anticipate the unique spread patterns associated with different variants.
By analyzing 9 million SARS-CoV-2 genetic sequences collected from 30 countries by the Global Initiative on Sharing Avian Influenza Data (GISAID), along with data on vaccination rates, infection rates, and other variables, Retsef Levi and his team explored the factors influencing viral transmission.
A predictive model leveraging machine learning was developed based on insights gleaned from this analysis. Within just one week post-diagnosis, the model can pinpoint 72.8% of the variant trends in each country that are likely to lead to over 1,000 cases per million people in the subsequent three months.
With two days of monitoring, this predictive accuracy escalates to 80.1%. Key predictors of a variant’s contagiousness include the initial surge of infections attributed to the variant, its evolving mutations, and the extent of genetic divergence from the predominant strain during the monitoring period.
The researchers suggest that this modeling approach could be extended to forecast the trajectory of other infectious diseases in the future.
