Genome-Wide Mutation Scoring for Machine-Learning-Based Antimicrobial Resistance Prediction

Abstract

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.

Validation and Application of Long-Read Whole-Genome Sequencing for Antimicrobial Resistance Gene Detection and Antimicrobial Susceptibility Testing

“…Predictive antimicrobial susceptibility testing from genome assemblies (WGS-AST) was determined by machine learning models as described previously ( 41 , 42 ). In brief, classification models were trained per species-antimicrobial pair on the AMR reference database ARESdb ( 24 ) and supplemented with rule-based classifiers to predict the susceptible or resistant AST interpretive category.…”

Automated antimicrobial susceptibility testing and antimicrobial resistance genotyping using Illumina and Oxford Nanopore Technologies sequencing data among Enterobacteriaceae

“…The model stacks combine extreme gradient boosting, elastic net regularized logistic regression (ENLR), and set covering machine models as well as rule-based post-processing routines. ARESdb models were trained on features derived from sequence motifs and variants ( Lüftinger et al, 2021b ; Májek et al, 2021 ). Ertapenem was used as a proxy for the prediction of resistance to carbapenems and was optimized from previous publications, to recognize non-CP CREs ( Májek et al, 2021 ) by recognizing changes in the ompK35 , ompK36, and homologs ( Doumith et al, 2009 ; Sugawara et al, 2016 ).…”

“…ARESdb models were trained on features derived from sequence motifs and variants ( Lüftinger et al, 2021b ; Májek et al, 2021 ). Ertapenem was used as a proxy for the prediction of resistance to carbapenems and was optimized from previous publications, to recognize non-CP CREs ( Májek et al, 2021 ) by recognizing changes in the ompK35 , ompK36, and homologs ( Doumith et al, 2009 ; Sugawara et al, 2016 ). WGS-AST was run for 60 species-antimicrobial pairs listed in Supplementary Table 1 .…”

Prediction of genome-wide imipenem resistance features in Klebsiella pneumoniae using machine learning

“…The k-mer as input feature could capture various types of genetic determinants with no prior knowledge and improve model interpretability since a k-mer feature represents a specificity gene [17][18][19][20][21][22], it is more effective for prediction and analysis of antibiotic resistance. The length of k-mer has a large effect on model interpretation and memory capacity.…”