An AI tool better detected meropenem-resistant Klebsiella in African and Thai data

The genomic tool was more sensitive than two established resistance databases and missed fewer resistant strains in testing.

Abstract editorial illustration accompanying the article: An AI tool better detected meropenem-resistant Klebsiella in African and Thai data

Editorial illustration — not from the study.

The team trained PredKlebAMR, a machine-learning system, on 5,942 K. pneumoniae genomes using 38 gene-based features. The system was designed to prioritize sensitivity, meaning it aimed to reduce cases in which resistant bacteria are incorrectly classified as susceptible.

On a held-out global test set, the tool had 91.92% accuracy, 94.67% sensitivity and 89.47% specificity. In independent comparisons, it had higher sensitivity and fewer very major errors—missed cases of resistance—than ResFinder 4.1 and AMRFinderPlus 4.2.7 across global, African and Thai datasets.

How the tool performed

PredKlebAMR identified meropenem-resistant K. pneumoniae with 95.45% sensitivity in the global cohort, compared with 90.91% for AMRFinderPlus and 86.36% for ResFinder. It recorded one very major error, compared with two for AMRFinderPlus and three for ResFinder.

In the African cohort, PredKlebAMR reached 95.56% sensitivity, compared with 91.11% for AMRFinderPlus and 82.2% for ResFinder, with two very major errors versus four and eight. In the Thailand cohort, sensitivity was 95.62%, compared with 92.5% and 63.12%, with seven very major errors versus 12 and 59. The improvement in sensitivity in the African comparison was statistically significant, with a McNemar’s test p-value of 0.041.

Important predictors included mutations linked to fluoroquinolone resistance, high-risk bacterial sequence types, KPC-2 and KPC-3 resistance genes, and changes to the OmpK35 and OmpK36 membrane proteins. The tool is available through an interactive R Shiny web application that provides a biological rationale for its predictions.

Why missed resistance matters

Carbapenem-resistant K. pneumoniae can limit treatment options, so missing a resistant strain during susceptibility profiling can have serious consequences for antibiotic decisions. A tool that prioritizes sensitivity could help identify more resistant strains in genomic testing, particularly in the African and Thai datasets examined here.

The findings show improved detection compared with two commonly used genomic resistance resources, but the study does not establish how using the tool would affect patient treatment or outcomes.

Evidence and limits

This was a machine-learning study based on bacterial genome datasets. PredKlebAMR was internally evaluated using a stratified 80/20 split and five-fold cross-validation, then compared with ResFinder 4.1 and AMRFinderPlus 4.2.7 in independent global, African and Thai cohorts from the NCBI database.

The abstract does not report the sizes or sampling details of the external cohorts. The evaluation measured prediction performance rather than clinical outcomes, and the study does not show whether the tool performs similarly on all K. pneumoniae populations or in prospective clinical testing. The reported statistical significance applied to the African sensitivity comparison, not necessarily to every comparison.

Human brain organoids classified Braille letters with up to 83% accuracy

Human forebrain organoids classified artificial tactile Braille signals after a sensor converted the touch data into electrical stimulation. Accuracy averaged 61% with one organoid and rose to 83% when responses from three organoids were combined.

1 hour ago·1 views

Biology

A peptide reduced persistent traumatic memories in mice

A study in mice found that a peptide targeting the memory-related protein CPEB3 reduced the persistence of traumatic memories and anhedonia, a loss of pleasure. The findings are an early genetic preclinical test of a possible approach for PTSD, not evidence of a treatment in people.

6 hours ago·5 views

Biology

In fruit flies, heat and pesticide stress cut fertility across generations

Acute heat and sublethal imidacloprid exposure interacted to reduce male fruit flies’ reproductive success more than either stressor alone in some conditions. Stress experienced by fathers also reduced offspring fertility, while a more diverse gut microbiome partly lessened the effects on males.

20 hours ago·5 views