MVPHI: a multi-view learning framework for predicting complex microbial interactions.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Bacteriophages (phages) regulate microbial communities and are crucial for microbiome turnover.
• Predicting phage-bacteria interactions (PBIs) and bacteria-bacteria interactions (BBIs) is vital for microbiome research.
• Current wet-lab methods for predicting interactions are costly and risky, necessitating computational alternatives.

Purpose of the Study:

• To develop a highly accurate and efficient computational model for predicting complex microbial interactions.
• To address the limitations of existing bioinformatics approaches in predicting PBIs and BBIs.

Main Methods:

• Proposed a multi-learning based model named MVPHI.
• Constructed a heterogeneous multi-attributed microbial network (MAMN) using pathogenic bacteria and phages.
• Utilized three distinct feature sets: statistical-view, textual-view, and topology-view for model training.

Main Results:

• MVPHI demonstrated superior performance over six variant models and eight baseline algorithms across seven benchmark datasets.
• Case studies and protein docking experiments confirmed the model's robustness and generalization capabilities.
• The model achieved high accuracy in predicting complex microbial interactions.

Conclusions:

• The MVPHI model shows significant potential for predicting novel PBIs and BBIs.
• This computational tool can provide valuable insights for phage screening and bacterial community research.
• MVPHI offers an efficient and reliable alternative to traditional experimental methods for studying microbial interactions.