New method for a near real-time identification of airborne bacteria

The new technique, based on Laser-Induced Fluorescence and machine learning, opens the door to faster and more accurate monitoring of environmental bioaerosols.It is now possible to identify bacteria present in the air almost instantly, without the need to collect samples or process them in the laboratory. The study, shows that a portable device can recognise different types of bacteria in aerosols by combining an ultraviolet laser that induces fluorescence in their components with an artificial intelligence system capable of interpreting that signal. Microorganisms present in the atmosphere—the so-called aerobiome—play a key role in ecosystems, climate and human health. However, the study has been limited by complex technical challenges: air samples usually contain extremely low amounts of bacterial DNA, which slows down sequencing-based approaches. Automated bioaerosol detectors have made significant progress in pollen identification, but until now there has been no effective method to distinguish microbial particles in real time, a limitation that hinders environmental surveillance and rapid response to biological threats or bioaerosol-related pollution events.To achieve this, the team adapted a commercial device known as Rapid-E, replacing its original laser with a 266 nm laser capable of exciting—i.e. inducing fluorescence in—compounds characteristic of bacteria. They then generated laboratory aerosols containing five bacterial species commonly found in urban environments, simulating particles that may be present in ambient air. The device analysed each particle by measuring how it scattered light and the type of fluorescence it emitted—a kind of optical “fingerprint” for each microorganism—and these data were fed into machine learning models trained to recognise species-specific patterns. The results show that the system can distinguish between bacterial and non-bacterial particles with an accuracy of 96.7%, and identify specific bacterial species with an average accuracy of 69.2%, a remarkable achievement given the small size and complexity of microbial particles. Extension is underway to other microorganisms such as fungi and giant viruses.