Capturing organism-specific knowledge on AMR for interpreting genomic data

Organism-specific interpretation of antimicrobial susceptibility testing (AST) data is standard in clinical microbiology, with rules regularly reviewed by expert committees of CLSI and EUCAST. EUCAST also maintains lists of expert rules for some species, including expected (intrinsic) resistance and expected susceptibility phenotypes, to guide clinical labs in deciding which drugs to test and whether/how to report them. We propose there is a similar need for systematic rules for the organism-specific interpretation of antimicrobial resistance (AMR) genotypes derived from pathogen whole-genome sequence data, as the presence of the same gene can have different functional implications in different organisms. For example, the oqxABR operon is core to the Klebsiella pneumoniae chromosome, where its expression is typically regulated to a level below the clinical breakpoint for fluoroquinolone resistance (ciprofloxacin MIC <0.25 mg/L). However, oqxABR can be mobilised from the K. pneumoniae chromosome into other species, where expression is regulated differently. Therefore, the presence of oqxAB in non-K. pneumoniae species typically results in clinical resistance to fluoroquinolones (i.e. ciprofloxacin MIC >0.5 mg/L), and so the functional implications of oqxAB detection are species-specific. Whilst there are bespoke solutions for AMR interpretation in specific organisms (e.g. the Kleborate tool for K. pneumoniae; Pathogenwatch AMR libraries for Neisseria gonorrhoeae and others; ResFinder for Salmonella, E. coli, and others), this complicates bioinformatic analyses and interpretation, and promotes fragmentation rather than consolidation of expertise. In collaboration with the AMRnet team at London School of Hygiene and Tropical Medicine, we are developing a novel data structure to capture and utilise organism-specific rules relating to functional interpretation of individual AMR determinants. We are also implementing new bioinformatics workflows to use this data structure to facilitate rapid interpretation of genome data for public health benefit. The new data structures are simple, designed to be curated and updated by microbiology experts without bioinformatics expertise, and can include PubMed identifiers to annotate rules with relevant supporting evidence. (see examples at https://github.com/katholt/orgspecAMR). In this project, the student will select a bacterial species for which AMR is of significant concern (e.g. species of Klebsiella, Enterobacter, Pseudomonas, Acinetobacter), and develop and test a rule-set for interpreting AMR genomic data from that organism. Aims will be to review the literature for evidence on genetic mechanisms for resistance and use the findings from the literature review to define organism-specific rules for this species that can be used to interpret raw AMRfinderplus output. Experience with genomics/bioinformatics analysis is useful but not essential. Sufficient knowledge of microbiology and molecular biology to understand the molecular basis of AMR mechanisms is required. This project is a collaboration with the London School of Hygiene and Tropical Medicine.