Computational prediction of inter-species relationships through omics data analysis and machine learning

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Antibiotic resistance poses a global threat to healthcare.
• Phage therapy, using viruses to kill bacteria, is a promising alternative to antibiotics.
• Efficiently matching therapeutic phages to specific bacteria is a significant challenge.

Purpose of the Study:

• To develop a computational model for predicting phage-bacterium interactions.
• To overcome the limitations of empirical testing for phage-bacterium pair identification.
• To accelerate the discovery of novel phage therapies.

Main Methods:

• Collected extensive phage-bacterium interaction data and genomic information.
• Extracted features from protein-protein interactions and genomic sequences.
• Trained machine learning models using these features.

Main Results:

• Developed predictive models with approximately 90% accuracy (F1-score, sensitivity, specificity, accuracy).
• Validated model performance using 10-fold cross-validation on a test set.
• Demonstrated the potential of genomic data for predicting phage-bacterium interactions.

Conclusions:

• Machine learning can significantly accelerate the identification of interacting phage-bacterium pairs.
• This approach offers a powerful tool for managing multi-drug resistant infections.
• Computational prediction can streamline phage therapy development.