Szeged researchers develop machine learning method to track bacterial infections

Researchers at the Szeged Biological Research Centre (SZBK), in collaboration with the Hungarian Centre for Excellence in Molecular Medicine (HCEMM), have developed a new machine learning-based method to analyze bacterial infections at the single-cell level. The system, named MALVINA (Machine Learning-Based Virulence Interaction Analysis), allows for the systematic evaluation of how bacteria enter human cells, how they multiply, and the extent of DNA damage they cause.

Existing methods often lack the ability to perform single-cell assessments, which limits the study of how different bacterial strains interact or how drug treatments affect pathogen virulence. During testing, the system analyzed over 1.3 million individual human colon epithelial cells infected with four different fluorescently labeled E. coli strains, including harmless laboratory strains and disease-associated strains that produce DNA-damaging toxins.

According to lead author Benc Bognár, the method provides detailed insights by distinguishing whether a small number of bacteria enter many cells or if a large quantity accumulates in just a few cells. The system achieved an 86 percent accuracy rate in distinguishing between infected and uninfected cells. The study has been published in the journal Nature Communications.