A Comparative Study of Machine Learning Models for Identification of Antiviral Peptides Using Various Encoded...

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Viral infections pose significant global health threats, necessitating novel therapeutic strategies.
• Antiviral peptides (AVPs) show promise, but their identification is often slow and resource-intensive.
• Machine learning (ML) offers a powerful approach to accelerate the discovery of potential AVPs.

Purpose of the Study:

• To develop and evaluate an ML model for predicting effective antiviral peptides from protein sequences.
• To compare the performance of various ML algorithms and feature encoding methods for AVP prediction.
• To establish a reliable computational tool for identifying potential therapeutic antiviral peptides.

Main Methods:

• Eight ML algorithms were assessed, focusing on a Light Gradient Boosting Machine (LGBM) model.
• Three distinct feature-encoding techniques were employed for protein sequence representation.
• Model performance was evaluated using accuracy, precision, recall, F1-score, and AUC metrics.

Main Results:

• The LGBM model, utilizing combined feature encodings, achieved the highest performance.
• The best model demonstrated 98% accuracy, 97% precision, 98% recall, 98% F1-score, and an AUC of 1.00.
• The proposed approach outperformed existing models by approximately 2-3% in accuracy.

Conclusions:

• The developed LGBM model is highly effective and reliable for predicting antiviral peptides.
• This computational approach significantly accelerates the identification of potential AVPs for therapeutic development.
• The findings hold considerable value for pharmaceutical research and academic endeavors in combating viral diseases.