Seq-BEL: Sequence-Based Ensemble Learning for Predicting Virus-Human Protein-Protein Interaction.

Area of Science:

• Virology
• Bioinformatics
• Computational Biology

Background:

• Infectious diseases pose a significant global health challenge, necessitating a deeper understanding of viral infection mechanisms.
• Predicting virus-human protein-protein interactions (PPIs) is crucial for controlling highly infectious viral diseases.
• Existing prediction methods struggle with data imbalances, lack of non-interactive protein pairs, and poor generalization for viral proteins.

Purpose of the Study:

• To develop an effective computational method for predicting potential virus-human PPIs.
• To address the limitations of traditional supervised learning and ensemble methods in PPI prediction.
• To enhance the understanding of viral pathogenesis through accurate interaction prediction.

Main Methods:

• A Sequence-Based Ensemble Learning (Seq-BEL) method was proposed, utilizing amino acid sequences of human and viral proteins.
• Seq-BEL calculates protein features and similarities based on known virus-human PPI networks.
• A scoring system combines these features and similarities to predict potential PPIs.

Main Results:

• Seq-BEL demonstrated high success in predicting potential virus-human PPIs.
• The method outperformed current state-of-the-art techniques in prediction accuracy.
• Seq-BEL exhibited robust performance for novel human and viral proteins, indicating good generalization ability.

Conclusions:

• The Seq-BEL method is a robust and effective tool for predicting virus-human PPIs.
• Its strong generalization capability makes it valuable for identifying interactions involving new proteins.
• This approach contributes to understanding viral infection mechanisms and developing control strategies.