PHIAF: prediction of phage-host interactions with GAN-based data augmentation and sequence-based feature fusion

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Phage therapy presents a promising alternative to antibiotics for bacterial infections.
• Identifying phage-host interactions (PHIs) is crucial for understanding phage infection mechanisms and advancing phage therapy.
• Computational methods offer a cost-effective and efficient approach to PHI identification compared to traditional wet experiments.

Purpose of the Study:

• To propose a novel computational method, PHIAF, for predicting phage-host interactions (PHIs).
• To enhance PHI prediction accuracy by addressing data scarcity and improving feature representation.
• To provide an interpretable model for predicting PHIs to guide phage therapy development.

Main Methods:

• Utilized a generative adversarial network (GAN)-based data augmentation module to generate pseudo PHIs and alleviate data scarcity.
• Implemented a sequence-based feature fusion approach, integrating features from both DNA and protein sequences.
• Incorporated an attention mechanism to weigh the contributions of different sequence-derived features, enhancing model interpretability.

Main Results:

• PHIAF demonstrated superior performance over existing state-of-the-art methods in computational experiments, achieving AUC of 0.88 and AUPR of 0.86 via 5-fold cross-validation.
• Ablation studies confirmed the positive impact of data augmentation, feature fusion, and the attention mechanism on prediction performance.
• Case study identified four novel PHIs with high PHIAF scores, subsequently verified by recent literature.

Conclusions:

• PHIAF is an effective and promising computational tool for predicting phage-host interactions.
• The method's components, including GAN-based data augmentation and attention-based feature fusion, significantly contribute to its predictive power.
• PHIAF has the potential to accelerate the discovery and development of phage therapy applications.