A machine learning model built with data from thousands of patients with wound infections and urinary tract infections identified the factors that contribute to antibiotic resistance in recurrent infections. The authors of this work say a patient’s history of infections and antibiotic treatments can be used with patient demographic data to predict which candidate antibiotics would likely prevent a recurrent infection. “Machine learning recommendation systems such as the one presented [here] have the potential to substantially improve patient outcomes and could play a major role in mitigating antibiotic resistance,” write Jean-Baptiste Lagagne and Mary Dunlop in a related Perspective. In many cases, bacterial infections – including common urinary tract infections (UTIs) and wound infections – are seeded from bacteria from a patient’s microbiota. Treatment of these infections often involves using a wide range of antibiotics. Serious infections are often evaluated for antibiotic susceptibility, which guides the use of a particular drug. However, while the initial treatment may clear the infection, it’s thought that antibiotic use may pave the way for resistant strains to arise and replace the previous susceptible strain. Thus, infections initially diagnosed as antibiotic-susceptible and treated as such can reoccur and become life-threateningly drug-resistant. Using a large longitudinal dataset of more than 200,000 UTIs and wound infections and the associated patients’ microbiome profiles, Mathew Stracy and colleagues looked for incidences where initial antibiotic treatment was not effective. To better understand why some infections subsequently gained resistance in this group, they carried out genomic sequencing of patient bacteria in individuals who experienced early UTI recurrence, providing a detailed view of the strains and species of the original infection compared with the ones that caused it to recur. Resistance-gaining recurrences were caused by strain replacement not by point mutations in the originally infecting strain, they found. “This analysis reveals an underappreciated path to reinfection, with the original species being treated and eliminated but with the treatment ultimately setting the stage for other resistant strains to emerge,” write Lagagne and Dunlop in their Perspective. Stracy and team then used resulting data to develop a machine learning model that predicts the risks of a pathogen gaining resistance to a particular antibiotic at the individual-patient level.
Minimizing treatment-induced emergence of antibiotic resistance in bacterial infections
Article Publication Date