Predicting pathogen evolution and immune evasion in the age of artificial intelligence

Area of Science:

• Virology
• Genomics
• Epidemiology
• Artificial Intelligence

Background:

• Viral pathogens evolve genomically to evade public health interventions during epidemics and pandemics.
• Current strategies respond to viral variants post-detection, causing delays in effective responses.
• Predicting harmful viral mutations before they emerge is a critical, long-standing objective.

Purpose of the Study:

• To review recent advancements in forecasting viral evolution, particularly during pandemics.
• To highlight the role of artificial intelligence (AI) and deep learning in predicting viral mutations.
• To discuss adaptable pipelines for RNA virus evolution applicable across different species.

Main Methods:

• Leveraging AI, specifically deep learning architectures and language models (LM).
• Integrating diverse data types: genomic, epidemiologic, immunologic, and biological information.
• Developing and adapting predictive pipelines, initially for SARS-CoV-2, with broader RNA virus applicability.

Main Results:

• AI and deep learning show significant promise in forecasting viral evolution.
• The COVID-19 pandemic accelerated the development of predictive viral evolution tools.
• Methods are increasingly adaptable for predicting mutations in various RNA viruses.

Conclusions:

• AI-driven forecasting of viral evolution is a powerful tool for pandemic preparedness.
• Proactive identification of harmful viral mutations enhances public health response.
• Deep learning and language models offer novel avenues for predicting pathogen adaptation.