MoDPepInt

MoDPepInt predicts binding partners for modular protein domains (SH2, SH3, and PDZ) using domain-specific machine-learning models.

Key Features:

• SH2PepInt: Predicts interactions for 51 human SH2 domains using single-domain models.
• SH3PepInt: Predicts interactions for 69 human SH3 domains using single-domain models.
• PDZPepInt: Predicts interactions for 226 PDZ domains across multiple species using 43 multidomain models.
• SVM-based modeling: Uses support vector machines with polynomial, Gaussian, and graph kernels to model complex non-linear amino-acid residue interactions.
• Validation and benchmarking: Models were validated on manually curated datasets and exhibit competitive performance relative to other approaches.

Scientific Applications:

• PPI prediction for modular domains: Identifies potential binding partners for SH2, SH3, and PDZ domains to support studies of protein–protein interactions.
• Signaling mechanism analysis: Supports elucidation of molecular mechanisms in cellular signaling and disease by predicting domain-mediated interactions.
• Cross-species PDZ studies: Enables comparative analysis of PDZ-domain interactions across multiple species using multidomain models.

Methodology:

Support vector machines (SVMs) with polynomial, Gaussian, and graph kernels are applied to single-domain and multidomain models; model validation used manually curated datasets.