Rapid discovery of novel prophages using biological feature engineering and machine learning

Area of Science:

• Microbiology
• Bioinformatics
• Genomics

Background:

• Prophages, integrated into bacterial genomes, are crucial for understanding bacterial biology.
• The vast diversity of prophages makes detection via sequence similarity challenging, leading to many unidentified phages.
• Current detection methods are limited by reliance on sequence similarity and speed.

Purpose of the Study:

• To introduce a novel, fast, and generalizable machine learning method for prophage discovery.
• To overcome the limitations of sequence similarity-based detection methods.
• To enhance the identification of previously undiscovered prophages.

Main Methods:

• A novel machine learning approach utilizing feature space for prophage detection.
• Reanalysis of publicly available marine viromes and single-cell genomes.
• Comparison with current state-of-the-art prophage detection tools.

Main Results:

• The developed method identified significantly more phages compared to existing tools.
• The approach demonstrated notable speed improvements over current methods.
• Consistent enhancement of bacteriophage discovery across different datasets.

Conclusions:

• The feature-based machine learning method substantially improves prophage discovery.
• This advancement facilitates the exploration of novel bacteriophage biology.
• The method offers a new starting point for identifying diverse prophages.