Streptococcus pneumoniae (the pneumococcus) is an important global pathogen. It is the leading bacterial cause of pneumonia, a disease for which there are >100 million cases and 800,000 deaths in children under five annually. Pneumococci are also among the top six pathogens causing deaths attributable to antimicrobial resistance (AMR). Pneumococcal serotypes targeted by current vaccines often carry more AMR determinants, and by eliminating these from a population, AMR can often decrease. Genomics has been used successfully in high-income countries to measure pneumococcal AMR. However, there is a paucity of genomic data from pneumococci in the Asia-Pacific. As AMR and antimicrobial use are high in this region, existing pipelines and databases may be less accurate. Here, we explore the ability of currently available bioinformatic tools (pipelines and databases) to accurately predict AMR using isolates from vaccine impact studies in Lao PDR and Mongolia.
Pneumococci (n=560) isolated from pneumonia carriage study samples collected from Laos (n=348) and Mongolia (n=222) were tested phenotypically using a combination of disc diffusion and E-tests and using CLSI breakpoints. Pneumococcal isolates were sequenced (2x150bp) using the NovaSeq platform (Illumina) and analysed for AMR determinants using tools that are either customisable for pneumococci (n=3; CDC-developed pipeline, AMIMA and WamR) or are general AMR tools (n=5; ARG-ANNOT, CARD, MEGARes, AMRFinderPlus and ResFinder). Overall agreement between phenotypic methods and antimicrobials included in each tool was measured by a prevalence-adjusted bias-adjusted kappa analysis, with >80% agreement considered acceptable.
Preliminary analysis of three custom pneumococcal tools showed strong agreement with phenotypic results for non-beta-lactam antimicrobials. In contrast tools for non-penicillin beta-lactams (e.g. ceftriaxone in n=2 custom tools) performed poorly. General AMR databases performed well but don’t include predictions for beta-lactam resistance as well as several other antimicrobials. Further investigation is needed into some discrepancies which may indicate novel resistance determinants. These results suggested that a genomics-based approach to understanding pneumococcal AMR in the Asia-Pacific is potentially feasible but has some limitations.