Predicting the spread of SARS-CoV-2 variants: An artificial intelligence enabled early detection

Area of Science:

• Virology
• Epidemiology
• Computational Biology

Background:

• SARS-CoV-2 rapidly mutates into infectious variants, causing multiple global infection waves.
• Existing epidemiological models predict pandemic trajectories but do not incorporate variant-specific genetic data.
• Understanding variant spread is crucial for effective public health responses.

Purpose of the Study:

• To develop a predictive model for the spread of new SARS-CoV-2 variants.
• To integrate variant-specific genetic characteristics with epidemiological data for improved forecasting.
• To identify emerging variants likely to cause future infection waves.

Main Methods:

• Analysis of 9.0 million SARS-CoV-2 genetic sequences from 30 countries.
• Identification of temporal patterns associated with variants causing significant infection waves.
• Development of a machine-learning risk assessment model for variant-specific spread prediction.

Main Results:

• The model can predict emerging variants likely to cause new infection waves up to 3 months in advance.
• High predictive accuracy was achieved: 86.3% AUC after 1 week and 90.8% AUC after 2 weeks.
• Identified temporal characteristic patterns of SARS-CoV-2 variants driving infection waves.

Conclusions:

• Integrating genetic and epidemiological data enhances the prediction of variant-specific spread.
• The developed model offers early warning for potential new COVID-19 waves.
• The methodology can be applied to predict variants of other infectious viruses.