Innovative Alignment-Based Method for Antiviral Peptide Prediction.

Area of Science:

• Biochemistry
• Computational Biology
• Drug Discovery

Background:

• Antiviral peptides (AVPs) offer a vital strategy against viral infections and drug resistance.
• Traditional laboratory-based AVP discovery is costly and time-consuming.
• Efficient computational methods are needed to accelerate AVP identification.

Purpose of the Study:

• To develop and validate novel computational models for AVP discovery.
• To assess the performance of these models against existing AVP prediction tools.
• To create a comprehensive dataset of antiviral sequences.

Main Methods:

• Development of five supervised multi-query similarity search models (MQSSMs) within the StarPep toolbox.
• Rigorous testing and validation using diverse AVP datasets.
• Benchmarking MQSSMs against 14 public machine learning and deep learning AVP predictors.

Main Results:

• The developed MQSSMs demonstrated robustness and reliability across datasets.
• The top model, M13+, achieved high accuracy (0.969) and Matthew's correlation coefficient (0.71).
• MQSSMs outperformed existing predictors in efficiency and accessibility, and a comprehensive antiviral sequence dataset was compiled.

Conclusions:

• MQSSMs are effective alignment-based tools for screening large datasets for novel AVP discovery.
• These models present a computationally efficient and accessible approach to antiviral drug development.
• The study advances the field by providing superior predictive models and a valuable dataset for AVP research.