Integrating imaging and omics with machine learning approaches to infer antimicrobial susceptibility in gram-negative bacteria

Thesis

Antimicrobial resistance is a major global health issue, which threatens the efficacy of antimicrobial treatments. Many efforts have been made to study the genetic mechanisms of resistance and adaptation of bacteria under antimicrobial treatments. However, the impacts of these genetic factors on the bacterial adaptation have not been well understood. In this thesis, I holistically investigated the physiological and cytological adaptation of Gram-negative bacteria upon exposure to different antimicrobials. Particularly, proteomic and cellomic profiles along with genomic data of these bacteria were acquired, which provides insights into the molecular mechanism of the adaptation. I found a high prevalence of resistant pathogenic Escherichia coli isolated from children with sepsis or diarrhoeal disease. Although antimicrobial exposure generally reduced the growth bacteria, azithromycin accelerated the bacterial growth at sub-inhibitory concentrations. Additionally, tetracycline- resistant isolates were found to possess a fitness advantage over the susceptible isolates. Furthermore, a variety of machine learning models were trained on the proteomic and cellomic profiles, resulting in high accuracy of resistance identification. The models also found strong association of resistance to different antimicrobials with the proteomic and cellomic profiles. The machine learning models suggested that exposure to ciprofloxacin exerted a significant perturbation on bacterial membrane and the expression of various transmembrane proteins, which generates novel hypotheses for further experiments to confirm. Alternatively, I utilised a high-content imaging platform to observe the cytological adaptation of Salmonella enterica serovar Typhimurium upon 24-hour of ciprofloxacin i exposure. Strong association of the adaptation with the exposure time and the antimicrobial concentrations was found. I also found that resistant isolates behaved differently compared to the susceptible isolate even without exposure to the antimicrobial. In conclusion, further holistic investigations of antimicrobial resistance should be conducted with the aids of machine learning approaches to provide in-depth understanding on the mechanism of bacterial adaptation, contributing to the fight against antimicrobial resistance.

Authors: Tran, TA

• DOI: 10.5287/ora-exgpyny59
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: machine learning; ciprofloxacin; antimicrobial resistance; azithromycin; high-content imaging; Escherichia coli; tetracycline; Salmonella Typhimurium; Enterobacteriaceae
• Subjects: Machine learning; Artificial intelligence; Microscopy; Bacteriology; Clinical medicine