Scientists have developed a neural network which has discovered a clutch of new insect repellent molecules, many of them more powerful than DEET, the World's current gold standard.
University of Florida researcher Alan Kratizky and his colleagues began by developing an analytical computer tool that was capable of learning. Next the team fed into the system the chemical structures of 150 previously-tested insect repellent molecules together with how effective those substances were at warding off bugs. This way the model could spot the structures that were consistently present in chemicals that were good repellents.
Next they tested how well the system had learned by feeding in a further 50 chemical compounds and comparing the predictions from the model with how well those chemicals had been shown to work in real experiments. Once the lesson was over the system was put through its paces and used to screen a family of similar chemicals, called N-acylpiperidines, that had never been tested as insect repellents.
The model pinpointed 34 promising candidates, which the team then synthesised and tested by adding them to strips of muslin. The cloth strips were then placed on the hands of volunteers and exposed to hungry female mosquitoes whilst the team totted up how many bites occurred in the space of one minute.
All of the chemicals were effective, but some exceedingly so, to the extent that they were two to three times better than DEET. The safety of these chemicals will need to be evaluated before bush walkers can start slapping them on, but undoubtedly this is a big breakthrough, especially given that DEET has never been matched despite first being made in 1953.