Zeugodacus fruit flies (Diptera: Tephritidae) host preference analysis by machine learning-based approaches

Abstract

Detecting the host preference of highly polyphagous and economically significant pest species of fruit flies (Diptera; Tephritidae) is important for identifying their species status, their management in orchards and the international trade of fruits and vegetables. In the current study, three fruit fly species Zeugodacus tau, Z. signata, and Z. cucrbitae, (Diptera: Tephritidae) were evaluated for their oviposition preference among three host fruits: pumpkin, cucumber, and bitter gourd. The investigation was conducted under choice conditions in the laboratory. Fruit fly species and host fruits were used as input/predictive variables whereas, oviposition preference, number of pupae, weight of pupae, adult emergence, and sex ratio were used as output/response variables to test the host preference through an Artificial Neural Network ANN/machine learning (ML) algorithms. ANN-based on a Multi-Layer Perceptron (MLP) model and decision tree-based Random Forest (RF) models were employed. Results revealed that Z. tau preferred pumpkin > cucumber > bitter gourd in order, Z. cucurbitae preferred bitter gourd > pumpkin > cucumber in order and Z. signata also preferred pumpkin but followed by bitter gourd and cucumber for oviposition. The specific host preferences observed in both Z. tau and Z. signata suggest that they may not be distinct species but rather closely related siblings. These findings highlight host preference as a marker for species delimitation. Moreover, the machine learning (ML) tools, provide better prediction in identifying host preference than statistical methods. These results are discussed in the context of the importance of studying host preferences for fruit flies’ species delimitation, their management, and international trade of fruits and vegetables.