B-134 Evaluation of marker gene-based in silico antimicrobial resistance (AMR) prediction tools

Abstract Background Monitoring and surveillance of antimicrobial resistance (AMR) is important procedure in clinically patient management and epidemiologically public health. Conventionally such culture based tools as disk diffusion method or broth dilution method for antibiotic susceptibility test has been used. While recently such culture independent approaches as PICRUSt2, Tax4Fun, or MicFunPred have been tried based on predictive functional profiling using 16S rRNA marker gene, evaluation of the tools regarding AMR is little. Methods A total of 20 E. coli strains (Carbapenem resistant (CRE) positive: 10, CRE negative: 10) were used. The AMR phenotype was based on Vitek2 (biomerieux). DNA was extracted from the 20 strains and 16S rRNA (V3-V4 region) and shotgun sequencing were done. The bioinformatic pieplines were QIIM2 for 16S rRNA and MetaPhlAn4 for shotgun. Functional prediction tools were PICRUSt2, Tax4Fun, and MicFunPred for 16S rRNA and AMRFinderPlus for shotgun. The presence/absence of 25 KEGG numbers regarding AMR in PICRUSt2, Tax4Fun, and MicFunPred were compared to shotgun AMR profiles. The F1 scores were calculated according to each 16S marker gene based prediction tools using confusion matrix. Results A total of 14 class of antibiotics including carbapenem were analyzed. The F1 scores of 16S predictive functional profilers regarding AMR were 0.22 for Tax4Fun, 0.12 for PICRUSt2, and 0.08 for MicFunPred. Conclusion While Tax4Fun showed the highest F1 score among three 16S predictive functional profilers, the F1 scores were generally low. The strength of Tax4Fun is the incorporation of user-defined data and the enhanced accuracy of the predictions is expected through incorporation of AMR-specific data in engine training.