Development and validation of a MidInfraRed Spectroscopy approach for Aedes albopictus age grading

Development and validation of a MidInfraRed Spectroscopy approach for Aedes albopictus age grading M. Foti1, M. Micocci1, M.P. Betancourth2, I. Casas2, B. Caputo1, F. Baldini2 and A. Della Torre1 1SAPIENZA University of Rome, Public Health and Infectious Diseases, Piazzale Aldo Moro 5, 00185 Rome, Italy, 2University of Glasgow, School of Biodiversity, One Health, and Veterinary Medicine, University of Glasgow, Glasgow G12 8QQ, United Kingdom; alessandra.dellatorre@uniroma1.it Quantifying mosquito population age-structures is crucial to assess the dynamics of mosquito-borne-diseases and the effectiveness of vector control interventions. This is especially the case when transmission of arboviruses is not endemic and the assessment cannot be based on changes in human cases. However, determining the age of a mosquito vector is strongly hampered by the need to rely on labour-intensive ovarian dissections by well trained personnel. Many efforts have been placed in recent years to overcome this major constraint, particularly for major tropical vector species. One promising approach is based on mid-infrared spectroscopy (MIRS) and machine learning (ML), which can quickly and accurately determine the age of female Anopheles African malaria vector species based on the measurement of biochemical changes in the cuticle over time. Here, we show the results obtained by applying MIRS-ML for age determination of adult Aedes albopictus females and males from laboratory and semifield settings. Field collected Aedes albopictus eggs were reared to adulthood either under laboratory or semi-field conditions. Adult females were regularly blood-fed and allowed to lay eggs. Males and females at different physiological states were collected every 3 days until day-36 under laboratory conditions, and every 3 consecutive days (from 1−3 to 31−33 day old) in case of semi-field adults. Spectra from individual mosquitoes was acquired by Attenuated Total Reflection FT-IR spectroscopy using a Bruker ALPHA II spectrometer between 4000 and 400 cm−1 with 4 cm−1 resolutions. The dataset was split into training and test sets (80−20%). Several machine learning algorithms were tested, and logistic regression (LR) was selected for optimisation based on the accuracy of the predictions. Under laboratory conditions, LR model predicted the mosquito age of both male and female mosquitoes with 95% accuracy, ranging from 86% to 100%, depending on the age-group. We are currently analysing spectra of semi-field reared mosquitoes, where preliminary results suggest that age-grading is possible also under these more variable conditions. These results support the high potential of MIRS-ML as the first non-morphological approach to accurately and cost-effectively assess the age structure of Ae. albopictus. Further studies are needed to confirm its reliability on predicting field-collected specimens.