Integrating developmental traits, morphometrics and machine learning for sex classification of Aedes aegypti...

Aedes aegypti (Linnaeus, 1762), the primary vector for dengue, Zika, chikungunya and yellow fever, requires precise sex separation for the effective application of the sterile insect technique. This study integrates developmental biology, morphometric analysis and machine learning to differentiate sexes in Ae. aegypti pupae and adults. We evaluated developmental parameters across six generations of the Old Balara strain. Morphometric data from pupae and adults, including pupal cephalothorax dimensions, pupal length, pupal weight, adult wing length and adult thorax length, were collected to assess their correlations. Pupal morphometrics (cephalothorax dimensions and pupal length and weight) were used to train nine ML models for pupal sex classification. Males pupated significantly earlier (5.61 ± 0.10 days) than females (6.40 ± 0.16 days). In the absence of mating, females lived longer (22.63 ± 0.42 days) than males (20.13 ± 0.34 days). After mating, longevity declined, with females showing a 10.34% reduction (2.34 days) and males a 32.94% reduction (6.63 days). Across all morphometric parameters (cephalothorax length, width and perimeter in mm; cephalothorax area in mm²; pupal length in mm; pupal weight in mg; adult wing length in mm; thorax length in mm), females were significantly larger than males. Pupal cephalothorax perimeter and area strongly correlated with adult thorax and wing lengths (R² up to 0.88), indicating their predictive value for adult size. Logistic Regression, Support Vector Machines and Naïve Bayes achieved 100% accuracy in pupal sex classification on the test set. Integrating morphometric analysis with machine learning significantly enhances sex sorting precision for the sterile insect technique, improving mosquito control efficiency by ensuring only males are released. Future research should explore incorporating image-based data and advanced machine learning techniques to further refine these methods across diverse mosquito populations.