Evaluation of WGS based tools in predicting antibiotic resistance in Pseudomonas aeruginosa

WGS is expected to revolutionize clinical diagnostic microbiology in the near future. However, its use as an antibiotic resistance prediction tool is still a subject of debate. Data are still lacking on the concordance between antibiotic resistance genotype and phenotype in many bacterial species. The aim of this work is to investigate for the concordance between phenotypic antimicrobial susceptibility testing and WGS-based resistance prediction in a genetically and phenotypically diverse collection of Ps.aeruginosa isolates. Three WGS-based bioinformatics tools were evaluated for the accuracy of resistance prediction for ciprofloxacin, levofloxacin and amikacin in 87 genetically diverse isolates of Ps.aeruginosa that include both sensitive and resistant isolates. Study results show that WGS-based tools do not perform well in predicting resistance for studied aminoglycosides and quinolones agents. Although high sensitivity rate was observed for some commonly known resistance-associated mutations, false positive rate (Type 1 eror) was very high using some predictive tools where most sensitive isolates were genomically predicted as resistant. We conclude that the use of WGS as a clinically applied predictive tool for antimicrobial resistance is not yet well proved in all bacterial species and further studies are still required in that respect. The set of commonly known resistance associated genes and mutations appears to be insufficient to account for antimicrobial resistance in Ps.aeruginosa.