Machine learning prediction accuracies ranged from 75.6% (P aeruginosa and ceftazidime; 90/119 correctly predicted) to 98.1% (P aeruginosa and amikacin; 105/107 correctly predicted).
When dealing with multidrug-resistant pathogens, every minute counts and any way to speed up the prediction of antimicrobial resistance (AMR) is key to reducing morbidity and mortality.
Infections with multidrug-resistant (MDR) Pseudomonas aeruginosa are increasing worldwide, and the organism is especially prevalent in health care-associated settings. But rapid AMR predictions could help with providing optimal care to patients.
Investigators in Texas have developed a new algorithm, variant mapping and prediction of AMR (VAMPr) using publicly available whole genome sequencing and antibiotic susceptibility testing (AST) data. In research presented at
To test it, investigators sequenced a validation cohort of contemporary P aeruginosa bloodstream isolates and performed AST. They then calculated an accuracy score around predicting AMR for various P aeruginosa-drug combinations.
VAMPr comprised a total of 3393 bacterial isolates (83 P aeruginosa isolates included) from 9 species, which contained AST data for 29 antibiotics.
“14,615 variant genotypes were identified within the dataset and 93 association and prediction models were built,” investigators explained in the abstract. “120 [P aeruginosa] bloodstream isolates from cancer patients were included for analysis in the validation cohort. ~15% of isolates were carbapenem resistant and ~20% were quinolone resistant.”
Machine learning prediction accuracies ranged from 75.6% (P aeruginosa and ceftazidime; 90/119 correctly predicted) to 98.1% (P aeruginosa and amikacin; 105/107 correctly predicted) for drug-isolate combinations where > 100 isolates were available. Machine learning also correctly identified known variants that strongly predicted resistance to various classes of antibiotics.
“Machine learning predicted AMR in P aeruginosa across a number of antibiotics with high accuracy,” investigators concluded. “Given the genomic heterogeneity of [P aeruginosa], increased genomic data for this pathogen will aid in further improving prediction accuracy across all antibiotic classes.”
The study, Machine Learning Approaches to Predicting Resistance in Pseudomonas aeruginosa, was presented in an oral abstract session on Friday, October 4, 2019, at IDWeek in Washington, DC.
