CoMPHI: a novel composite machine learning approach utilizing multiple feature representation to predict hosts of...

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Phage therapy offers a promising alternative to antibiotics for treating bacterial infections, particularly those caused by antibiotic-resistant superbugs.
• A significant challenge is identifying specific bacterial hosts for the large number of uncharacterized phages.

Purpose of the Study:

• To develop a robust computational framework for predicting phage-host interactions.
• To enhance the identification of suitable phages for therapeutic applications.

Main Methods:

• Introduction of the Composite Model for Phage Host Interaction (CoMPHI).
• Integration of alignment-based methods with machine learning, utilizing nucleotide and protein sequence features.
• Inclusion of alignment scores from phage-phage, phage-host, and host-host comparisons.

Main Results:

• CoMPHI achieved high performance metrics, including Area Under the ROC Curve (AUC-ROC) of 94-96.7% and accuracies of 92.3-95.1% across various taxonomic levels.
• Incorporating alignment scores improved model performance by 6-8%.
• Ablation studies confirmed that combining sequence features and alignment data significantly boosted prediction accuracy.

Conclusions:

• CoMPHI provides a comprehensive and accurate framework for predicting phage-host interactions.
• The model advances computational tools, potentially accelerating the clinical application of phage therapy.