Fig. 3

Summary of the experimental design and sample size of laboratory-reared (a) and wild-caught (b) Anopheles scanned to determine the accuracy of NIRS to predict Anopheles-Plasmodium infectious status according to preservation method. In the laboratory, the model was trained to predict Plasmodium infection status of An. coluzzii using 139 An. coluzzii mosquitoes (69.5 uninfected and 69.5 infected) per replication. This model was validated with 68 Anopheles and used to test 68 Anopheles (34 uninfected and 34 infected). To predict mosquito infection status in the field using ethanol as preservative, the model was trained using 25 An. coluzzii mosquitoes (12.5 uninfected and 12.5 infected) per replication. Validating and testing the model were realized with 12.5 Anopheles mosquitoes per data subset. With silica gel, the model was trained using 18 An. coluzzii mosquitoes (9 uninfected and 9 infected) per replication. Validating and testing the model were realized with 9 Anopheles mosquitoes per data subset