Starts 14 Jun 2018 11:00
Ends 14 Jun 2018 12:00
Central European Time
QLS Seminar: Vectors of Chagas disease - determination of species by infrared spectroscopy and machine learning
ICTP
Central Area, 2nd floor, old SISSA building, Via Beirut 2
Chagas disease, caused by the parasite Trypanosoma cruzi, is widespread in Latin America, where the disease remains one of the major public health problems, with an estimation of 8 million infected people and 10 000 deaths per year. This condition is mostly transmitted by insects, known as kissing bugs, belonging to the Triatominae family (Hemiptera), which are obligate haematophagous insects all their life. More than 150 species have been described in the world. While the majority lives in the wild, some of them are highly associated to the human beings, living in or around the dwellings. They typically stay hidden in the wall or roof cracks of homes, going out at night to feed on human blood. The correct determination of the species involved in the transmission of the disease is crucial to develop efficient control strategies. This can be achieved by keys of determination for adult stages, or by molecular technics. Both technics are time and/ or money consuming, showing the needs of new identification methods, especially for nymphal instars. These last years, various publications demonstrated the potential of infrared spectroscopy in insect taxonomy. Bolivia is a highly endemic country for Chagas disease. The main vector, Triatoma infestans, lives on about 60% of the territory. In total, 17 species of triatomines are reported in the country, among them Triatoma sordida and Triatoma guasayana which are reported as secondary vectors. These two species are sympatric and morphologically similar, and so they are difficult to discriminate. The goal of this study was to develop a classification model, using living nymphal and adult stage of these three species. 1293 spectra were taken in the invisible and near-infrared range. Different models were built, using different pre-processing methodologies of the spectra, and different types of feature selection. The performance of each model was evaluated for each species. After their comparison, the best model was tested on a different set of specimens where it showed a global accuracy of 97% (95-99%), an F1 score greater than 0.95 and a specificity greater than 0.94. This result shows that using infrared spectroscopy is a good strategy to predict the triatomine species. It is the first investigation to report the ability to identify juvenile instars, and moreover with a single model together with the adult stage.