Identification of viral dose and administration time in simulated phage therapy occurrences

Abstract

AbstractThe rise in multidrug-resistant bacteria has sprung a renewed interest in applying phages as antibacterial, a procedure Western practitioners eventually abandoned due to several downfalls, including poor understanding of the dynamics between phages and bacteria. A successful phage therapy needs to account for the loss of infective virions and the multiplication of the hosts. The parameters critical inoculation size (VF) and failure threshold time (TF) have been introduced to assure that the viral dose (vϕ) and administration time (tϕ) would lead to an effective treatment. The problem with the definition of VF and TF is that they are non-linear equations with two unknowns; thus, their solution is cumbersome and not unique. The current study used machine learning in the form of a decision tree algorithm to determine ranges for the viral dose and administration times required to achieve an effective phage therapy. Within these ranges, a Pareto optimal solution of a multi-criterial optimization problem (MCOP) provides values leading to effective treatment. The algorithm was tested on a series of microbial consortia that described allochthonous invasions (the outgrowing of a species at high cell density by another species initially present at low concentration) to inhibit the growth of the invading species. The present study also introduced the concept of ‘mediated phage therapy’, where targeting a booster bacteria might decrease the virulence of a pathogen immune to phagial infection. The results demonstrated that the MCOP could provide pairs of vϕ and tϕ that could effectively wipe out the bacterial target from the considered micro-environment. In summary, the present work introduced a novel method for investigating the phage/bacteria interaction that could help increase the effectiveness of phage therapy.Author summaryPhage therapy is a treatment that can help fight infections with bacteria resistant to antibiotics. However, several phage therapy application have failed, possibly because phages were administered at the wrong time or in insufficient amounts. The present study implemented a machine learning protocol to correctly calculate the administration time and viral load to obtain effective phage therapy. Four simulated microbial consortia, including one case where the pathogen was not directly a phage’s host, were employed to prove the procedure’s concept. The results demonstrated that the procedure is suitable to help the microbiologists to instantiate an effective phage therapy and clear infections.