Identification of New Mobile Antibiotic Resistance Genes using Metagenomic Abundances

Abstract

Antibiotic resistance is an increasing concern worldwide, as it compromises the treatments of infectious diseases and their use in many medical procedures. This project aims to identify antibiotic resistance genes (ARGs) that are spreading among bacterial populations but have not yet been detected in known pathogens. To achieve this, ARG sequences were aligned against a large bacterial database of genomes. The resulting matches were filtered for relevance and analysed using correlation-based methods. Several strategies for handling zero counts in the data were tested and different correlation metrics were evaluated. Ultimately, a zero-inflated negative binomial model combined with Pearson’s weighted correlation metrics was selected as it was the most effective approach. The method found matches between genes and bacterial hosts, that were then compared to already known hosts found in the literature. A small group of genes was selected based on the weak taxonomic relation of their identified hosts to the known hosts. Detailed analysis of these genes showed that the method had been successful in identifying several potentially mobile ARGs that appear to be in the early stages of spreading into non-pathogenic bacterial hosts.