Machine-learning prediction of resistance to sub-inhibitory antimicrobial concentrations from Escherichia coli genomes

Abstract

Escherichia coli is an important cause of bacterial infections worldwide, with multidrug resistant strains incurring substantial costs on human lives. Besides therapeutic concentrations of antimicrobials in healthcare settings, the presence of sub-inhibitory antimicrobial residues in the environment and in the clinics selects for antimicrobial resistance (AMR), but the underlying genetic repertoire is less well understood. We used machine-learning to predict the population doubling time and growth yield of 1,432 genetically diverse E. coli expanding under exposure to three sub-inhibitory concentrations of six classes of antimicrobials from single nucleotide genetic variants, accessory gene variation and the presence of known AMR genes. We could predict cell yields in the held-out test data with an average correlation (Spearmans {rho}) of 0.63 (0.32 - 0.90 across concentrations) and cell doubling time with an average correlation of 0.47 (0.32 - 0.74 across concentrations), with moderate increases in sample size unlikely to improve predictions further. This points to the remaining missing heritability of growth under antimicrobials exposure being explained by effects that are too rare or weak to be captured unless sample size is dramatically increased, or by effects other than those conferred by the presence of individual SNPs and genes. Predictions based on whole genome information were generally superior to those based only on known AMR genes, and also accurate for AMR resistance at therapeutic concentrations. We also pinpointed genes and SNPs determining the predicted growth and thereby recapitulated the known AMR determinants. Finally, we estimated the effect sizes of resistance genes across the entire collection of strains, disclosing growth effects for known resistance genes for each strain. Our results underscore the potential of predictive modelling of growth patterns from genomic data under sub-inhibitory concentrations of antimicrobials, although the remaining missing heritability poses an issue for achieving the accuracy and precision required for clinical use. ImportancePredicting bacterial growth from genome sequences is important not only for a rapid characterization of strains in clinical diagnostic applications but for the identification of novel targets for drug discovery. Previous studies examined the relationship between bacterial growth and genotype in mutant libraries for laboratory strains, yet no study has so far examined the prediction power of genome sequences for bacterial growth in natural strains. In this study, we used a high throughput phenotypic assay to measure bacterial growth of a systematic collection of natural Escherichia coli strains and then employed machine learning models to predict bacterial growth from genomic data under non-therapeutic sub-inhibitory concentrations of antimicrobials that are common in nonclinical settings. Our results revealed a moderate to strong correlation between predicted and actual values for different antimicrobials concentrations. Furthermore, the quantified effect of resistance genes on bacterial growth indicate these genes are still effective at sublethal antimicrobial concentrations.