PhagePickr: A bacteria-centric computational tool for designing resistance-aware phage cocktails

Abstract

As antibiotic resistance poses a major threat to global health, phage therapy offers an alternative to antibiotic treatments in the face of multidrug-resistant bacteria. However, host resistance to phages is also well-documented. Current computational tools for phage cocktail design do not explicitly address the evolution of phage resistance, let alone through the profiling of bacterial receptors whose variability drives much of phage resistance. We introduce PhagePickr, a computational pipeline for the automated design of phage cocktails intended to reduce the risk of host resistance. Unlike other tools, PhagePickr selects phages based on bacterial surface receptor similarity and prioritizes genomic diversity with the aim of reducing cross-resistance. The tool uses receptor annotations to identify phenotypically similar hosts via a nearest-neighbor algorithm and uses multiple sequence alignments to ensure genomic diversity in the final cocktail. We evaluated the utility of PhagePickr on ESKAPE pathogens and two understudied bacterial species. The cocktails included candidate phages selected using bacterial receptor similarity and phage genomic divergence, comprising both lytic phages with confirmed therapeutic potential and novel candidates from similar species. We demonstrate the tool's utility in generating cocktails and its capacity to scale as current databases are updated. PhagePickr provides a novel bacteria-centric framework for resistance-aware cocktail design by exploring shared phenotypes.