PrMFTP

PrMFTP predicts multi-functional therapeutic peptides (MFTPs) from peptide sequences to identify peptides with multiple biological activities for drug discovery.

Key Features:

• Multi-Head Self-Attention Mechanism: Captures intricate dependencies within peptide sequences to assess how different sequence regions contribute to functionality.
• Multi-Scale Convolutional Neural Network (CNN): Extracts and learns informative features from peptide sequences at multiple spatial scales to recognize patterns associated with multi-functionality.
• Bi-Directional Long Short-Term Memory (Bi-LSTM): Models sequential dependencies in peptide sequences by considering both forward and backward contexts.
• Class Weight Optimization Algorithm: Adjusts class importance during training to mitigate label imbalance and improve prediction of underrepresented MFTP classes.

Scientific Applications:

• Therapeutic peptide discovery: Predicts candidate MFTPs for identification of novel therapeutic agents.
• Candidate prioritization: Assists in ranking promising peptides for downstream experimental validation.
• Multi-target drug design: Supports design and selection of peptides with multiple biological activities to target diverse disease mechanisms.

Methodology:

Combines multi-head self-attention, multi-scale CNN, and Bi-LSTM to learn sequence features, with a class weight optimization algorithm applied during training to address label imbalance.