Developing machine learning classifiers for carbapenem resistance prediction in Klebsiella pneumoniae using Google's
João V O Cruz1, Felipe G Sacoda1, Vitor H Moreau2
1Post-Graduate Program in Biotechnology, Institute of Health Sciences, Federal University of Bahia, Canela, 40110-100 Salvador, Bahia, Brazil.
Journal of Microbiological Methods
|January 2, 2026
Abstract:
We developed neural network models trained with images from Klebsiella pneumoniae genomes, encoded via Chaos Game Representation, to predict carbapenem resistance. Both real (n = 168) and synthetic AI-generated (n = 200) images were used. The best-performing model, among the 72 developed, achieved up to 87.5 % accuracy on generalization datasets.


