PhageDPO: A machine-learning based computational framework for identifying phage depolymerases

Area of Science:

• Microbiology
• Bioinformatics
• Biotechnology

Background:

• Bacteriophages (phages) are abundant viruses with uncharacterized proteins.
• Depolymerases (DPOs) are phage enzymes degrading bacterial polysaccharides, with biotechnological potential.
• Existing DPO identification tools lack robustness.

Purpose of the Study:

• To develop PhageDPO, a reliable tool for predicting DPOs in phage genomes.
• To enhance the identification of DPO enzymes for research and applications.

Main Methods:

• Training a Support Vector Machine (SVM) model using a comprehensive dataset of DPO-related domains and validated DPOs.
• Utilizing seven specific DPO-related domains for model training.
• Validating the model with literature data and newly generated experimental data.

Main Results:

• PhageDPO achieved high performance metrics: 96% test accuracy, 97% recall, 94% precision, and 96% F1-score.
• The model demonstrated robust predictive capability for DPOs in phage genomes.
• PhageDPO offers a user-friendly interface and superior performance compared to existing tools.

Conclusions:

• PhageDPO is a highly accurate and reliable tool for identifying depolymerases in bacteriophage genomes.
• The tool facilitates the discovery of novel DPO enzymes for biotechnological applications, including pathogen control.
• PhageDPO enhances accessibility and effectiveness in phage-derived enzyme research.