Tiny sensors similar to alcohol breathalysers could be used to detect bacterial infections and identify antimicrobial resistance (AMR) in bodily fluids, according to an opinion article published in the journal Cell Biomaterials. The authors, a team of engineers, microbiologists and machine learning specialists based in Switzerland, said the technology could lead to affordable, rapid diagnostic tools that would enhance treatment decisions and help address AMR.
“One of the biggest drivers of AMR is that we lack rapid diagnostics,” said Dr Andreas Güntner, senior author and a mechanical and process engineer at ETH Zurich, who collaborated on the project with Dr Catherine Jutzeler, Dr Thomas Kessler, Professor Emma Slack and Professor Adrian Egli.
The researchers proposed bypassing conventional multi-step laboratory procedures that can take many hours, days or even weeks, and instead make use of handheld chemical sensors which would be capable of delivering results in seconds or minutes.
Historically, clinicians have used smell to detect certain infections. For example, Pseudomonas aeruginosa can produce a sweet, grape-like odour, whereas Clostridium infections are associated with a putrid smell. These scents are caused by volatile organic compounds (VOCs), small molecules released by bacteria and other organisms.
The authors argued that VOCs in blood, urine, faeces or sputum could be measured by specially designed sensors, offering a reliable proxy for the presence of infection. Similar approaches have been used to detect contaminants such as methanol in alcoholic beverages and to monitor air quality.
“We have already developed and commercialised something similar for detecting methanol. Now, we are trying to transfer this technology to more complex situations,” explained Güntner.
Because even closely related bacterial strains can emit distinct patterns of VOCs, the approach could help to distinguish those AMR strains. Laboratory studies have already shown that methicillin-resistant Staphylococcus aureus (MRSA) and its non-resistant counterpart can be identified based on their VOC profiles.
Developing clinically viable sensors, however, presents a significant challenge. The concentration of VOCs is extremely low and detecting them accurately requires materials with high sensitivity.
“Imagine a room full of a billion tiny blue balls [but] only one red one,” said Güntner. “You must identify that single red ball – and do so within [a few] seconds.”
The authors envisaged that future devices would incorporate arrays of sensors with varying binding properties, built using metal oxides, polymers, carbon nanotubes, and graphene derivatives. Filters would be required to exclude irrelevant compounds, such as human-derived or ubiquitous VOCs.
Machine learning would be instrumental in sensor development, enabling algorithms to identify the minimal set of VOCs necessary to detect bacterial species and assess resistance or virulence factors.
Ultimately, the researchers said their aim was to translate advances in chemical sensing and machine learning into practical diagnostic tools that could be used in everyday medical settings with minimal training.
“We hope this will improve patient outcomes and support antibiotic stewardship,” said Güntner.
For further reading please visit: 10.1016/j.celbio.2025.100125