Skip to main content
ARS Home » Southeast Area » Gainesville, Florida » Center for Medical, Agricultural and Veterinary Entomology » Chemistry Research » Research » Publications at this Location » Publication #330874

Research Project: Insect, Nematode, and Plant Semiochemical Communication Systems

Location: Chemistry Research

Title: Machine learning for characterization of insect vector feeding

item Willett, Denis
item George, Justin
item WILLETT, NORA - Princeton University
item STELINSKI, LUKASZ - University Of Florida
item Lapointe, Stephen

Submitted to: PLoS Computational Biology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/20/2016
Publication Date: N/A
Citation: N/A

Interpretive Summary: Insect vectors acquire and transmit pathogens causing infectious diseases through probing on host tissues and ingesting host fluids. By connecting insects and their food source via an electrical circuit, computers, using machine learning algorithms, can learn to recognize insect feeding patterns involved in pathogen transmission. In addition, these machine learning algorithms can show us novel patterns of insect feeding and uncover mechanisms that lead to disruption of pathogen transmission. While we use these techniques to help save the citrus industry from a major decline due to an insect-transmitted bacterial pathogen, such intelligent monitoring of insect vector feeding will engender advances in disrupting transmission of pathogens causing disease in agriculture, livestock, and human health.

Technical Abstract: Insects that feed by ingesting plant and animal fluids cause devastating damage to humans, livestock, and agriculture worldwide, primarily by transmitting pathogens of plants and animals. The feeding processes required for successful pathogen transmission by sucking insects can be recorded by monitoring voltage changes across an insect-food source feeding circuit. The output from such monitoring has traditionally been examined manually, a slow and onerous process. We taught a computer program to automatically classify previously described insect feeding patterns involved in transmission of the pathogen causing citrus greening disease. We also show how such analysis contributes to discovery of previously unrecognized feeding states and can be used to characterize plant resistance mechanisms. This advance greatly reduces the time and effort required to analyze insect feeding, and should facilitate developing, screening, and testing of novel intervention strategies to disrupt pathogen transmission affecting agriculture, livestock and human health.