Emvirus: An embedding-based neural framework for human-virus protein-protein interactions prediction

Area of Science:

• Computational biology
• Virology
• Machine learning

Background:

• Human-virus protein-protein interactions (PPIs) are central to viral pathogenesis, exemplified by SARS-CoV-2 spike protein binding to ACE2.
• Experimental identification of PPIs is resource-intensive, necessitating advanced computational approaches.

Purpose of the Study:

• To develop and evaluate an advanced computational method for predicting human-virus PPIs, including those involving SARS-CoV-2.
• To improve the accuracy and efficiency of identifying critical viral-host interactions.

Main Methods:

• An embedding-based neural network framework, Emvirus, integrating Convolutional Neural Networks (CNN) and Bi-directional Long Short-Term Memory (Bi-LSTM) units was developed.
• The model utilizes sequence-based features, moving beyond traditional manually extracted characteristics.
• Cross-viral experiments were performed to assess the model's generalization capabilities.

Main Results:

• E ভাইরাস demonstrated superior prediction accuracy compared to existing methods that rely on manually extracted features.
• The model's performance indicates its effectiveness in identifying human-virus PPIs.
• Cross-viral validation confirmed the model's robust generalization ability.

Conclusions:

• E ভাইরাস offers a powerful and accurate deep learning approach for predicting human-virus PPIs.
• This method can accelerate the identification of therapeutic targets for viral infections.
• The framework's generalization capability suggests broad applicability across different viruses.