What if a dangerous bacterium could be spotted before it ever made anyone sick? That is the aim of PathogenFinder2, an AI tool developed by researchers at the Technical University of Denmark (DTU) with international partners, and described in the journal Bioinformatics.
Until now, knowing whether a bacterium could cause disease often required complex lab experiments, or a comparison with bacteria already known. With an entirely new species, those methods quickly reach their limits. PathogenFinder2 instead relies on models that “read” proteins, a bit like learning a language, to recognize patterns associated with the ability to cause disease. It was trained on more than 21,000 bacterial genomes.
According to the researchers, the tool could analyze bacteria found in wastewater, in healthy people, or in animals. That could help prepare tests, vaccines, or treatments earlier. They are clear, however: an AI prediction is not a diagnosis. It is a monitoring tool.
Foreseeing is also a way of caring. This kind of research does not promise a world without illness, but it may give us a little more time to protect the most vulnerable.