Separation of species and age by REIMS. Mosquitoes (total n = 540) were raised to three different age groups (1 day, 5–6 days, 14–15 days) for each of the three species (An. coluzzii, An. gambiae s.s, An. arabiensis, see text). The specimens were killed by dehydration and stored at room temperature with desiccant for 1–1.5 weeks prior to analysis. The samples were used to build two models: one separating the three species (a) and one separating the three age groups (b). Data was processed using PC-LDA in Offline Model Builder (using 100 PCs) (i) and R (using 235 PCs), latter visualised for each linear discriminant (LD 1 and LD 2) separately in form of kernel density (ii) and scatter plots (iii). The data matrix, exported from the Offline Model Builder software, was additionally analysed using the random forest algorithm in R; 70% of the samples were used for model building, 30% as test samples. The classification results are depicted as a bar graph, showing percentages of correctly and wrongly identified samples (iv). Depicted are the average values of 10 random forest repeat runs ± the standard error of the mean, with the range of accuracy values that were achieved in brackets
