Analysis of near infrared spectra for age-grading of wild populations of Anopheles gambiae

International audienceBackground : Understanding the age-structure of mosquito populations, especially malaria vectors such asAnopheles gambiae,is important for assessing the risk of infectious mosquitoes, and how vector controlinterventions may impact this risk. The use of near-infrared spectroscopy (NIRS) for age-grading has beendemonstrated previously on laboratory and semi-field mosquitoes, but to date has not been utilized onwild-caught mosquitoes whose age is externally validated via parity status or parasite infection stage. Inthis study, we developed regression and classification models using NIRS on datasets of wildAn. gambiae(s.l.) reared from larvae collected from the field in Burkina Faso, and two laboratory strains. We comparedthe accuracy of these models for predicting the ages of wild-caught mosquitoes that had been scored fortheirparitystatusaswellasforpositivityforPlasmodiumsporozoites.Results : Regression models utilizing variable selection increased predictive accuracy over the more common full-spectrum partial least squares (PLS) approach for cross-validation of the datasets, validation, and independent test sets.Models produced from datasets that included the greatest range of mosquito samples (i.e. different sampling locationsand times) had the highest predictive accuracy on independent testing sets, though overall accuracy on these sampleswas low. For classification, we found that intramodel accuracy ranged between 73.5–97.0% for grouping of mosquitoesinto“early”and“late”age classes, with the highest prediction accuracy found in laboratory colonized mosquitoes.However, this accuracy was decreased on test sets, with the highest classification of an independent set of wild-caughtlarvae reared to set ages being 69.6%.Conclusions : Variation in NIRS data, likely from dietary, genetic, and other factors limits the accuracy of this techniquewith wild-caught mosquitoes. Alternative algorithms may help improve prediction accuracy, but care should be takento either maximize variety in models or minimize confounders