Figure 1

The ability of NIRS to predict laboratory-reared mosquitoes infectious with wild parasites. All models were trained on sporozoite positive and sporozoite negative laboratory reared mosquitoes using all the data presented in Table 1. (A) Receiver operating characteristic (ROC) curve illustrating the diagnostic ability of the best-fit model. Overall performance is given by the average area under the ROC curve (AUC). Figure illustrates the false positive and true positive rates achievable for different classification probability thresholds. A theoretical perfect diagnostic would be in the top left corner. Average ROC curve shown by the solid line with boxplots showing the variability for 50 randomizations of the training, validation and testing datasets (horizontal black line shows the median whilst the 25th/75th, 15th/85th and 5th/95th percentiles are shown by box edges, inner and outer whiskers, respectively). (B) Coefficient functions for the best fit model for each of the 50 dataset randomizations (grey lines) and the overall average (black line). (C) Histogram showing the predicted status of tested mosquitoes that were infectious (light blue colored bars) or uninfectious (green bars).Vertical solid black line indicates the best threshold for differentiating between infectious or uninfectious mosquitoes. Darker blue bars indicates where the two distributions overlap and show those mosquitoes misclassified—false negatives are shown to the left of the optimal classification threshold line and false positives to the right. Inset shows the confusion matrix illustrating the different error rates: true negative rate (tnr, specificity); false negative rate (fnr); false positive rate (fpr); and true positive rate (tpr, sensitivity).