Deep-AmPEP30: Improve Short Antimicrobial Peptides Prediction with Deep Learning

Area of Science:

• Biotechnology
• Computational Biology
• Drug Discovery

Background:

• Antimicrobial peptides (AMPs) offer a promising solution to combat multi-drug resistance.
• Short-length AMPs exhibit superior antimicrobial activity, stability, and reduced human cell toxicity.

Purpose of the Study:

• To develop and validate Deep-AmPEP30, a computational tool for predicting short-length AMPs (≤30 amino acids).
• To identify novel AMPs from genomic data for potential therapeutic applications.

Main Methods:

• Utilized a reduced amino acid composition feature set (PseKRAAC) and a convolutional neural network.
• Trained and evaluated the model on a balanced benchmark dataset of 188 samples.
• Screened the genome of *Candida glabrata* for potential AMPs.

Main Results:

• Deep-AmPEP30 achieved 77% accuracy, 85% AUC-ROC, and 85% AUC-PR on a benchmark dataset.
• Identified a 20-amino acid peptide (P3) with potent antibacterial activity comparable to ampicillin.
• The tool demonstrated superior performance compared to existing machine learning methods.

Conclusions:

• Deep-AmPEP30 is an effective prediction tool for identifying short AMPs from genomic sequences.
• The tool facilitates accelerated drug discovery for novel antimicrobial agents.
• The developed method is publicly accessible for sequence prediction and genome screening.