Machine learning helps identify chemicals that repel honey bees from pesticides

Researchers at the University of California, Riverside have utilized machine learning to identify chemical compounds that can repel honey bees from pesticide-treated crops, offering a potential solution to the declining bee population. Led by Anandasankar Ray, an expert in insect olfactory behavior, the interdisciplinary team tackled the challenge of finding scents that repel bees due to the complexity of their olfactory system.

The machine-learning model, trained on existing data and refined with new behavioral information, screened over 50 million compounds and identified about 130 potential bee repellents. In lab tests, all seven top-performing compounds reliably repelled bees from honeycombs without causing harm. This demonstrates the potential of machine learning in solving ecological problems and reducing pesticide exposure, which could contribute to bee population decline and colony collapse.