ACVPICPred: Inhibitory activity prediction of anti-coronavirus peptides based on artificial neural network

Area of Science:

• Computational biology
• Drug discovery
• Virology

Background:

• Peptides offer safe and effective coronavirus inhibition, crucial for developing new antiviral drugs.
• Accelerating the identification of active anti-coronavirus peptides (ACoVPs) is vital for combating coronavirus diseases.
• Existing methods for ACoVPs activity prediction require enhancement for accuracy and robustness.

Purpose of the Study:

• To develop ACVPICPred, a computational model for predicting ACoVPs inhibitory activity.
• To integrate sequence and structural information for improved prediction accuracy.
• To enhance model robustness using data augmentation techniques.

Main Methods:

• Utilized AlphaFold3 for peptide structural predictions.
• Employed bioinformatics tools for feature extraction from sequence and structure.
• Applied data augmentation (noise injection, SMOGN) to improve dataset robustness.
• Performed five-fold cross-validation for performance evaluation.

Main Results:

• The model achieved a Pearson correlation coefficient of 0.7668 and R² of 0.5880 on the training dataset.
• Models incorporating structural features demonstrated superior performance over sequence-only models.
• Data augmentation techniques enhanced overall model robustness.

Conclusions:

• ACVPICPred effectively predicts ACoVPs inhibitory activity by integrating sequence and structural data.
• Combined sequence and structural features yield more robust prediction results.
• The developed model and its findings contribute to accelerating anti-coronavirus drug development.