Prediction of Antigenic Distance in Influenza A Using Attribute Network Embedding

Area of Science:

• Virology
• Computational Biology
• Immunology

Background:

• Influenza viruses rapidly change antigenicity, hindering lasting human immunity and effective antiviral therapies.
• Dynamic tracking of influenza virus antigenic shifts is essential for developing targeted vaccines and treatments.

Purpose of the Study:

• To develop a novel quantitative prediction method for antigenic distance between influenza virus strains.
• To utilize attribute network embedding techniques for modeling influenza A virus H3N2 genetic and antigenic characteristics.

Main Methods:

• Constructed an antigenic network incorporating genetic and antigenic data of influenza A virus H3N2.
• Employed ProtVec (protein sequence distributed representation) as node attributes and antigenic distance as edge weights.
• Applied attribute network embedding techniques for quantitative prediction.

Main Results:

• Demonstrated a strong positive correlation between incorporating genetic features and prediction accuracy of antigenic distance.
• The developed model accurately tracks antigenic distance variations between vaccine and circulating influenza virus strains.
• Outperformed existing methods in predicting antigenic distances between influenza virus strains.

Conclusions:

• The novel quantitative prediction method effectively models influenza virus antigenic evolution.
• Integrating genetic information significantly enhances the accuracy of antigenic distance predictions.
• This approach offers a more comprehensive and accurate tool for monitoring influenza virus antigenic drift and informing vaccine development.