Learning from prepandemic data to forecast viral escape.

Area of Science:

• Virology
• Computational Biology
• Immunology

Background:

• Pandemic preparedness requires predicting viral mutations that escape host immunity for effective vaccine and therapeutic design.
• Current prediction methods are limited, relying on experimental data or current strain prevalence, hindering early pandemic response.

Purpose of the Study:

• To develop a generalizable computational framework, EVEscape, for predicting viral mutation escape potential.
• To enable early identification of concerning viral variants before extensive surveillance or experimental data is available.

Main Methods:

• EVEscape combines a deep learning model trained on historical sequences with biophysical and structural information.
• The framework quantifies the viral escape potential of mutations at scale.

Main Results:

• EVEscape, trained on pre-2020 sequences, accurately predicts SARS-CoV-2 pandemic variation, comparable to experimental methods.
• The framework demonstrates generalizability across diverse viruses, including influenza, HIV, Lassa, and Nipah viruses.

Conclusions:

• EVEscape provides a scalable tool for anticipating viral evolution and immune escape, crucial for proactive vaccine and therapeutic development.
• The framework offers continually updated escape scores and forecasts emerging strains for ongoing pandemic preparedness.