Rapid and non-destructive identification of Anopheles gambiae and Anopheles arabiensis mosquito species using Raman spectroscopy via machine learning classification models

Abstract

Abstract Background Identification of malaria vectors is an important exercise that can result in the deployment of targeted control measures and monitoring the susceptibility of the vectors to control strategies. Although known to possess distinct biting behaviours and habitats, the African malaria vectors Anopheles gambiae and Anopheles arabiensis are morphologically indistinguishable and are known to be discriminated by molecular techniques. In this paper, Raman spectroscopy is proposed to complement the tedious and time-consuming Polymerase Chain Reaction (PCR) method for the rapid screening of mosquito identity. Methods A dispersive Raman microscope was used to record spectra from the legs (femurs and tibiae) of fresh anaesthetized laboratory-bred mosquitoes. The scattered Raman intensity signal peaks observed were predominantly centered at approximately 1400 cm−1, 1590 cm−1, and 2067 cm−1. These peaks, which are characteristic signatures of melanin pigment found in the insect cuticle, were important in the discrimination of the two mosquito species. Principal Component Analysis (PCA) was used for dimension reduction. Four classification models were built using the following techniques: Linear Discriminant Analysis (LDA), Logistic Regression (LR), Quadratic Discriminant Analysis (QDA), and Quadratic Support Vector Machine (QSVM). Results PCA extracted twenty-one features accounting for 95% of the variation in the data. Using the twenty-one principal components, LDA, LR, QDA, and QSVM discriminated and classified the two cryptic species with 86%, 85%, 89%, and 93% accuracy, respectively on cross-validation and 79%, 82%, 81% and 93% respectively on the test data set. Conclusion Raman spectroscopy in combination with machine learning tools is an effective, rapid and non-destructive method for discriminating and classifying two cryptic mosquito species, Anopheles gambiae and Anopheles arabiensis belonging to the Anopheles gambiae complex.

Keywords

• Raman spectroscopy
• Machine learning
• Discriminant Analysis
• Support Vector Machine
• Logistic Regression
• Melanin