An integrative approach to DNA barcoding, geometric morphometrics, and machine learning for field identification of Culex mosquitoes (Diptera: Culicidae), with implications for vector-borne disease surveillance
4
Issued Date
2025-11-01
Resource Type
ISSN
0001706X
eISSN
18736254
Scopus ID
2-s2.0-105020039942
Journal Title
Acta Tropica
Volume
271
Rights Holder(s)
SCOPUS
Bibliographic Citation
Acta Tropica Vol.271 (2025)
Suggested Citation
Laojun S., Changbunjong T., Kamoltham T., Chaiphongpachara T. An integrative approach to DNA barcoding, geometric morphometrics, and machine learning for field identification of Culex mosquitoes (Diptera: Culicidae), with implications for vector-borne disease surveillance. Acta Tropica Vol.271 (2025). doi:10.1016/j.actatropica.2025.107885 Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/112914
Title
An integrative approach to DNA barcoding, geometric morphometrics, and machine learning for field identification of Culex mosquitoes (Diptera: Culicidae), with implications for vector-borne disease surveillance
Author's Affiliation
Corresponding Author(s)
Other Contributor(s)
Abstract
Culex mosquitoes are of considerable medical and veterinary importance, acting as vectors of arboviruses such as Japanese encephalitis, Rift Valley fever, and West Nile virus, as well as the filarial parasite Wuchereria bancrofti. Accurate identification of Culex species, however, remains challenging due to their close morphological similarity, frequent damage to field-collected specimens, and the limited availability of trained taxonomists. To address these challenges, this study employed an integrative framework combining DNA barcoding, wing geometric morphometrics (GM), and Random Forest (RF) to improve the identification of 12 Culex species (Cx. bicornutus, Cx. bitaeniorhynchus, Cx. brevipalpis, Cx. fuscocephala, Cx. gelidus, Cx. hutchinsoni, Cx. nigropunctatus, Cx. pseudovishnui, Cx. quinquefasciatus, Cx. sinensis, Cx. sitiens, and Cx. tritaeniorhynchus) in Thailand. DNA barcoding successfully validated the morphological identifications, with nucleotide sequences from representative specimens showing strong concordance with the GenBank and Barcode of Life Data Systems (BOLD) databases (≥96 %), confirming the reliability of morphological diagnoses. Complementarily, wing GM demonstrated stronger discriminatory power: Mahalanobis distance analysis revealed all species to be significantly different (p < 0.05), and a cross-validated reclassification test achieved 82.18 % performance with an adjusted total accuracy of 80 %. For field identification of unknown specimens, both Mahalanobis distance and RF produced comparable results, yielding very high accuracy (80 %–100 %) for eight species. Overall, the integration of DNA barcoding, wing GM, and machine learning offers a robust and practical framework for strengthening mosquito-borne disease surveillance. Nonetheless, as each method has distinct strengths and limitations, their application should be carefully adapted to specific epidemiological and operational contexts.