Researchers at INRAE have developed an artificial intelligence method to identify the sex pheromone of the lily moth, a destructive agricultural pest whose larvae feed on lily crops across Asia and Oceania. The collaborative team, which includes scientists from Université Côte d'Azur and Nanjing Agricultural University, also mapped the olfactory receptors that detect the pheromone. The findings appear in BMC Biology as part of the EXPLOR'AE program.
Sex pheromones drive insect mating behavior. Males detect these chemical signals at extremely low concentrations, sometimes from hundreds of meters away. For agricultural pest management, understanding these pheromone systems enables farmers to deploy targeted biocontrol strategies that disrupt mating without chemicals. France already uses pheromone-based approaches to protect crops. The lily moth poses a genuine threat to production because its caterpillars consume plant tissue directly.
Traditional identification of pheromones relies on labor-intensive chemical extraction and analysis. Scientists collect insects, extract volatile compounds, and test each one individually to determine which triggers behavioral responses in males. This process takes months or years. The INRAE team bypassed this bottleneck with AI-driven analysis.
The researchers applied machine learning algorithms to genomic and proteomic data from lily moth antennae. Olfactory receptors sit on sensory cells in antennae and bind to pheromone molecules. By analyzing receptor sequences and their expression patterns, the AI identified which receptors respond to specific chemical signals. The team then synthesized candidate pheromone molecules and tested whether they activated these receptors in controlled laboratory settings.
This computational-first approach proved effective. The researchers identified the lily moth's pheromone and its associated receptor system without relying solely on traditional extraction methods. The speed and precision of the method open pathways to decode pheromones in other pest species much faster than conventional approaches allow.
The implications extend beyond lily moths. Hundreds of insect pest species remain poorly characterized in terms of their chemical communication. Conventional pheromone identification for each species could take decades. The AI method compresses this timeline dramatically. Once researchers understand a pest's pheromone system, they design lures or disruptants that either attract insects to traps or confuse males into failing to locate mates. These biological controls reduce crop damage while minimizing pesticide use.
For France's agricultural sector, this matters considerably. Pheromone-based biocontrol strategies already protect vineyards, orchards, and vegetable crops. Expanding the toolkit to include newly identified pheromones addresses emerging pests and expands options for farmers seeking chemical-free protection. The lily moth represents one clear target; others will follow.
Limitations exist. The AI approach depends on high-quality genomic data from target species, which not all insects have. Additionally, identifying the pheromone receptor is distinct from validating field efficacy. Laboratory confirmation that a receptor binds a molecule does not automatically guarantee that deploying lures in orchards will reduce pest populations. Field trials remain essential. The INRAE team validated their findings experimentally, but scaling the method to dozens of species will require sustained research investment and access to genomic resources.
The work represents a bridge between computational biology and agricultural pest management. By combining machine learning with traditional testing, INRAE demonstrates how modern tools can accelerate solutions to age-old farming challenges. Future efforts will test whether this approach generalizes across diverse insect taxa and delivers measurable benefits in commercial crop protection scenarios.