iDNS3IP: Identification and Characterization of HCV NS3 Protease Inhibitory Peptides

TL;DR

This paper introduces a new computational tool to identify peptides that inhibit the HCV NS3 protease, which could help in developing better antiviral therapies.

Contribution

The first computational model specifically designed to identify HCV NS3 protease inhibitory peptides using machine learning and a web-based tool.

Findings

Machine learning models achieved 98.85% accuracy using SVM and 97.83% with RF in predicting NS3IPs.

The web-based tool iDNS3IP enables real-time prediction and includes a BLAST-based similarity search.

Feature space analyses showed distinguishable clustering between inhibitory and non-inhibitory peptides.

Abstract

Hepatitis C virus (HCV) infection remains a significant global health burden, driven by the emergence of drug-resistant strains and the limited efficacy of current antiviral therapies. A promising strategy for therapeutic intervention involves targeting the NS3 protease, a viral enzyme essential for replication. In this study, we present the first computational model specifically designed to identify NS3 protease inhibitory peptides (NS3IPs). Using amino acid composition (AAC) and K-spaced amino acid pair composition (CKSAAP) features, we developed machine learning classifiers based on support vector machine (SVM) and random forest (RF), achieving accuracies of 98.85% and 97.83%, respectively, validated through 5-fold cross-validation and independent testing. To support the accessibility of the strategy, we implemented a web-based tool, iDNS3IP, which enables real-time prediction of NS3IPs. In addition, we performed feature space analyses using PCA, t-SNE, and LDA based on AAindex descriptors. The resulting visualizations showed a distinguishable clustering between NS3IPs and non-inhibitory peptides, suggesting that inhibitory activity may correlate with characteristic physicochemical patterns. This study provides a reliable and interpretable platform to assist in the discovery of therapeutic peptides and supports continued research into peptide-based antiviral strategies for drug-resistant HCV. To enhance its flexibility, the iDNS3IP web tool also incorporates a BLAST-based similarity search function, enabling users to evaluate inhibitory candidates from both predictive and homology-based perspectives.

Figures