The prediction of antimicrobial resistance (AMR) based on genomic information can improve patient outcomes. Genetic mechanisms have been shown to explain AMR with accuracies in line with standard microbiology laboratory testing. To translate genetic mechanisms into phenotypic AMR, machine learning has been successfully applied. AMR machine learning models typically use nucleotide k-mer counts to represent genomic sequences. While k-mer representation efficiently captures sequence variation, it also results in high-dimensional and sparse data. With limited training data available, achieving acceptable model performance or model interpretability is challenging. In this study, we explore the utility of feature engineering with several biologically relevant signals. We propose to predict the functional impact of observed mutations with PROVEAN to use the predicted impact as a new feature for each protein in an organism’s proteome. The addition of the new features was tested on a total of 19,521 isolates across nine clinically relevant pathogens and 30 different antibiotics. The new features significantly improved the predictive performance of trained AMR models for Pseudomonas aeruginosa, Citrobacter freundii, and Escherichia coli. The balanced accuracy of the respective models of those three pathogens improved by 6.0% on average.

Genomic antimicrobial susceptibility testing (AST) has been shown to be accurate for many pathogens and antimicrobials. However, these methods have not been systematically evaluated for clinical metagenomic data. We investigate the performance of in-silico AST from clinical metagenomes (MG-AST). Using isolate sequencing data from a multi-center study on antimicrobial resistance (AMR) as well as shotgun-sequenced septic urine samples, we simulate over 2000 complicated urinary tract infection (cUTI) metagenomes with known resistance phenotype to 5 antimicrobials. Applying rule-based and machine learning-based genomic AST classifiers, we explore the impact of sequencing depth and technology, metagenome complexity, and bioinformatics processing approaches on AST accuracy. By using an optimized metagenomics assembly and binning workflow, MG-AST achieved balanced accuracy within 5.1% of isolate-derived genomic AST. For poly-microbial infections, taxonomic sample complexity and relatedness of taxa in the sample is a key factor influencing metagenomic binning and downstream MG-AST accuracy. We show that the reassignment of putative plasmid contigs by their predicted host range and investigation of whole resistome capabilities improved MG-AST performance on poly-microbial samples. We further demonstrate that machine learning-based methods enable MG-AST with superior accuracy compared to rule-based approaches on simulated native patient samples.