Nature Communications (Mar 2024)

Accurate and rapid antibiotic susceptibility testing using a machine learning-assisted nanomotion technology platform

  • Alexander Sturm,
  • Grzegorz Jóźwiak,
  • Marta Pla Verge,
  • Laura Munch,
  • Gino Cathomen,
  • Anthony Vocat,
  • Amanda Luraschi-Eggemann,
  • Clara Orlando,
  • Katja Fromm,
  • Eric Delarze,
  • Michał Świątkowski,
  • Grzegorz Wielgoszewski,
  • Roxana M. Totu,
  • María García-Castillo,
  • Alexandre Delfino,
  • Florian Tagini,
  • Sandor Kasas,
  • Cornelia Lass-Flörl,
  • Ronald Gstir,
  • Rafael Cantón,
  • Gilbert Greub,
  • Danuta Cichocka

DOI
https://doi.org/10.1038/s41467-024-46213-y
Journal volume & issue
Vol. 15, no. 1
pp. 1 – 14

Abstract

Read online

Abstract Antimicrobial resistance (AMR) is a major public health threat, reducing treatment options for infected patients. AMR is promoted by a lack of access to rapid antibiotic susceptibility tests (ASTs). Accelerated ASTs can identify effective antibiotics for treatment in a timely and informed manner. We describe a rapid growth-independent phenotypic AST that uses a nanomotion technology platform to measure bacterial vibrations. Machine learning techniques are applied to analyze a large dataset encompassing 2762 individual nanomotion recordings from 1180 spiked positive blood culture samples covering 364 Escherichia coli and Klebsiella pneumoniae isolates exposed to cephalosporins and fluoroquinolones. The training performances of the different classification models achieve between 90.5 and 100% accuracy. Independent testing of the AST on 223 strains, including in clinical setting, correctly predict susceptibility and resistance with accuracies between 89.5% and 98.9%. The study shows the potential of this nanomotion platform for future bacterial phenotype delineation.