FIRM-AVP

FIRM-AVP predicts antiviral peptides (AVPs) from amino acid sequences using a Feature-Informed Reduced Machine Learning approach to identify physicochemical and structural determinants of antiviral activity.

Key Features:

• Machine learning classification: Distinguishes AVPs from non-AVPs using a trained machine learning model.
• Physicochemical and structural features: Leverages features derived from amino acid sequence physicochemical properties and structural attributes.
• Secondary structure importance: Emphasizes secondary structure as a critical predictor of antiviral peptide activity.
• Feature-Informed Reduced Machine Learning: Implements a Feature-Informed Reduced Machine Learning approach to focus on informative features.
• Feature filtering and selection: Filters and selects the most relevant features for classification to improve prediction accuracy.
• Comparative performance: Demonstrates higher performance than models using all features and state-of-the-art single classifiers.

Scientific Applications:

• Antiviral drug discovery: Supports identification of candidate AVPs for development of antiviral therapeutics.
• Peptide engineering: Informs design and optimization of peptide sequences based on physicochemical and structural determinants of activity.
• AVP identification across systems: Facilitates discovery of novel antiviral peptides across diverse biological systems by highlighting predictive features.

Methodology:

Uses a machine learning-based classifier trained on features derived from physicochemical and structural properties of amino acid sequences, applies feature importance analysis with emphasis on secondary structure, and employs a Feature-Informed Reduced Machine Learning approach that filters and selects relevant features for classification between AVPs and non-AVPs.