Using machine-learning and mid infrared spectroscopy for rapid assessment of blood-feeding histories and parasite infection rates in field-collected malaria mosquitoes

Effective surveillance and control of malaria-transmitting mosquitoes require quantitative understanding of key biological attributes, namely: preferred blood –hosts of mosquitoes, proportions infected with parasites, survivorship, indoor/outdoor-biting behaviour and insecticide susceptibility.

Currently, identifying mosquito blood meals and plasmodium infections involve enzyme-linked immunosorbent assays (ELISA), or polymerase chain reactions (PCR), which are time consuming, laborious and require expensive reagents. However, advances in near-n=infrared spectroscopy (NIR) suggest potential for cheaper, quicker and non-invasive alternative for predicting age and species of mosquitoes, and detecting pathogens e.g Wolbachia and Zika virus in laboratory- infected Aedes.

Promisingly, mid-infrared (MIR) can provide even better accuracies since structural identities of bio-molecules are delineated at finer resolutions than in NIR bands. However, the spectroscopy-based methods have not been field-validated because entomologist lack comparative field samples of known attributes and advanced computational methods to process large spectral datasets.

We project will couple MIR-spectroscopy with machine-learning algorithms and validate them for rapid assessment of blood-feeding histories and infectiousness of field collected Anopheles arabiensis and Anopheles funestus, which dominate malaria transmission in Tanzania. I will calibrate the systems to identify different vertebrate blood meals in mosquito abdomen, and Plasmodium sporozoite in heads and thoraces. This field validation will enable scale-up of MIR based approaches, thereby significantly improving surveillance-responses and intervention monitoring.