Model evaluation for automated scoring of electropenetrography waveform data from mosquitoes

Authors: COOPER, ANASTASIA - Auburn University; HOPE, GABRIEL - Swarthmore College; ABBAS, MEHREZAT - Harvey Mudd College; KNELL, MILO - Harvey Mudd College; GUGLANI, DEVANSHI - Harvey Mudd College; VERNOOY, LILLIAN - Harvey Mudd College; TRAUL, ZACHARY - Harvey Mudd College; Mitzel, Dana; Backus, Elaine; REIF, KATHRYN - Auburn University

Submitted to: Nature Scientific Reports
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/5/2026
Publication Date: 7/19/2026
Citation: Cooper, A.M., Hope, G., Abbas, M., Knell, M., Guglani, D., Vernooy, L., Traul, Z., Mitzel, D.N., Backus, E.A., Reif, K.E. 2026. Model evaluation for automated scoring of electropenetrography waveform data from mosquitoes. Nature Scientific Reports. Available online:. https://doi.org/10.1038/s41598-026-57373-w.
DOI: https://doi.org/10.1038/s41598-026-57373-w

Interpretive Summary: Blood-feeding arthropods like mosquitoes are economically devastating vectors of pathogens for humans, companion animals, and livestock. Yet limited information on the feeding behaviors of blood-feeding vectors greatly impedes development of control measures for such pathogens. Electropenetrography (EPG) is a powerful tool for quantifying how arthropods interact with their hosts, by translating into waveforms the intricate feeding behaviors otherwise invisibly performed in opaque host tissues. However, manually labeling feeding behaviors on EPG waveforms is subjective, labor-intensive, and time-consuming. Machine-learning models for supervised classification have achieved widespread success in pattern recognition and have the potential to aid in automating and standardizing the EPG waveform labeling process. Performance of various common machine learning methods were compared for the EPG labeling task, using a representative waveform dataset from Culex tarsalis mosquitoes. Results from the top five models are also presented. A neural network based on the UNet architecture with attention layers was the top-performing model, achieving an overall accuracy of 89%. To our knowledge, this is the first attempt to apply machine learning to identify waveforms from a blood-feeding arthropod, rather than a plant-feeding hemipteran. The findings of this study will facilitate the development of an automated waveform identification tool that researchers can use to score waveforms from mosquitoes and other blood-feeding arthropods.

Technical Abstract: Electropenetrography (EPG) is a powerful tool for quantifying how arthropods interact with their hosts, but manually labeling feeding behaviors on EPG waveforms is subjective, labor-intensive, and time-consuming. Machine-learning models for supervised classification have achieved widespread success in pattern recognition within scientific contexts and have the potential to aid in automating and standardizing the waveform labeling process. Performance of various common machine learning methods were compared for the EPG labeling task, using a representative waveform dataset from Culex tarsalis mosquitoes. Results from the top five models are also presented. A neural network based on the UNet architecture with attention layers was the top-performing model, achieving an overall accuracy of 89% and a macro F1 score of 0.78. To our knowledge, this is the first attempt to apply machine learning to identify waveforms from a blood-feeding arthropod, rather than a plant-feeding hemipteran. Experiments indicate that previously proposed machine learning approaches for classifying hemipteran waveforms do not generalize well to waveforms from blood-feeding arthropods. This suggests that analyzing species with distinct waveform patterns may require differing machine learning approaches. The findings of this study will facilitate the development of an automated waveform identification tool that researchers can use to score waveforms from mosquitoes and other blood-feeding arthropods.