PSSP-MVIRT

PSSP-MVIRT predicts peptide secondary structure by integrating sequential, evolutionary, and hidden state information into a unified feature space using a multi-view fusion strategy and a hybrid CNN-Bi-GRU network.

Key Features:

• Multi-View Fusion Strategy: Integrates sequential, evolutionary, and hidden state information to create a comprehensive feature representation of peptide sequences.
• Hybrid Network Architecture: Combines convolutional neural networks (CNNs) with bidirectional gated recurrent units (Bi-GRUs) to extract local and contextual features from peptide sequences.
• Transfer Learning: Utilizes transfer learning to leverage pretrained representations and improve model training on limited peptide datasets.

Scientific Applications:

• Therapeutic Peptide Research: Predicts peptide secondary structures to support exploration of peptides as potential therapeutic agents.
• Structural Biology Studies: Provides secondary-structure predictions to inform analyses of peptide behavior and interactions.

Methodology:

Multi-view fusion of sequential, evolutionary, and hidden state information; feature extraction using CNNs and Bi-GRUs; and model training augmented with transfer learning.