ABSTRACT Antimicrobial resistance (AMR) continues to evolve as a major threat to human health and new strategies are required for the treatment of AMR infections. Bacteriophages (phages) that kill bacterial pathogens are being identified for use in phage therapies, with the intention to apply these bactericidal viruses directly into the infection sites in bespoke phage cocktails. Despite the great unsampled phage diversity for this purpose, an issue hampering the roll out of phage therapy is the poor quality annotation of many of the phage genomes, particularly for those from infrequently sampled environmental sources. We developed a computational tool called STEP3 to use the “evolutionary features” that can be recognized in genome sequences of diverse phages. These features, when integrated into an ensemble framework, achieved a stable and robust prediction performance when benchmarked against other prediction tools using phages from diverse sources. Validation of the prediction accuracy of STEP3 was conducted with high-resolution mass spectrometry analysis of two novel phages, isolated from a watercourse in the Southern Hemisphere. STEP3 provides a robust computational approach to distinguish specific and universal features in phages to improve the quality of phage cocktails, and is available for use at http://step3.erc.monash.edu/. IMPORTANCE In response to the global problem of antimicrobial resistance there are moves to use bacteriophages (phages) as therapeutic agents. Selecting which phages will be effective therapeutics relies on interpreting features contributing to shelf-life and applicability to diagnosed infections. However, the protein components of the phage...

The component parts of bacteriophage virions accurately defined by a machine-learning approach built on evolutionary features

10.1101/2021.02.28.433281