PHERI - Phage Host Exploration pipeline.

Antibiotic resistance is becoming a common problem in medicine, food, and industry, with multi-resistant bacterial strains occurring in all world regions. One of the possible future solutions is the use of bacteriophages in therapy. Bacteriophages are the most abundant form of life in the biosphere, so we can highly likely purify a specific phage against each target bacterium. The identification and consistent characterization of individual phages was a common form of phage work and included determining bacteriophages host-specificity. With the advent of new modern sequencing methods, there was a problem with the detailed characterization of phages in the environment identified by metagenome analysis. The solution to this problem may be to use a bioinformatic approach in the form of prediction software capable of determining a bacterial host based on the phage whole-genome sequence. The result of our research is the machine learning algorithm based tool called PHERI. PHERI predicts suitable bacterial host genus for purification of individual viruses from different samples. Besides, it can identify and highlight protein sequences that are important for host selection. PHERI is available at https://hub.docker.com/repository/docker/andynet/pheri. The source code for the model training is available at https://github.com/andynet/pheri_preprocessing, and the source code for the tool is available at https://github.com/andynet/pheri.