# 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.

## Key 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.