Evaluation of parameters affecting performance and reliability of machine learning-based antibiotic susceptibility testing from whole genome sequencing data

Prediction of antibiotic resistance phenotypes from whole genome sequencing data by machine learning methods has been proposed as a promising platform for the development of sequence-based diagnostics. However, there has been no systematic evaluation of factors that may influence performance of such models, how they might apply to and vary across clinical populations, and what the implications might be in the clinical setting. Here, we performed a meta-analysis of seven large Neisseria gonorrhoeae datasets, as well as Klebsiella pneumoniae and Acinetobacter baumannii datasets, with whole genome sequence data and antibiotic susceptibility phenotypes using set covering machine classification, random forest classification, and random forest regression models to predict resistance phenotypes from genotype. We demonstrate how model performance varies by drug, dataset, resistance metric, and species, reflecting the complexities of generating clinically relevant conclusions from machine learning-derived models. Our findings underscore the importance of incorporating relevant biological and epidemiological knowledge into model design and assessment and suggest that doing so can inform tailored modeling for individual drugs, pathogens, and clinical populations. We further suggest that continued comprehensive sampling and incorporation of up-to-date whole genome sequence data, resistance phenotypes, and treatment outcome data into model training will be crucial to the clinical utility and sustainability of machine learning-based molecular diagnostics. Author Summary Machine learning-based prediction of antibiotic resistance from bacterial genome sequences represents a promising tool to rapidly determine the antibiotic susceptibility profile of clinical isolates and reduce the morbidity and mortality resulting from inappropriate and ineffective treatment. However, while there has been much focus on demonstrating the diagnostic potential of these modeling approaches, there has been little assessment of potential caveats and prerequisites associated with implementing predictive models of drug resistance in the clinical setting. Our results highlight significant biological and technical challenges facing the application of machine learning-based prediction of antibiotic resistance as a diagnostic tool. By outlining specific factors affecting model performance, our findings provide a framework for future work on modeling drug resistance and underscore the necessity of continued comprehensive sampling and reporting of treatment outcome data for building reliable and sustainable diagnostics.