DIFERENCIAÇÃO DE FÊMEAS DE LUTZOMYIA CRUZI E LUTZOMYIA LONGIPALPIS (DIPTERA: PSYCHODIDAE: PHLEBOTOMINAE) POR ESPECTROSCOPIA DE INFRAVERMELHO

• Autor(a) principal: MATHEUS EUGÊNIO PORTO BARBOSA
• Data de Publicação: 2024
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFMS
• Texto Completo: https://repositorio.ufms.br/handle/123456789/8513
• topic: Lutzomya cruzi; Lutzomya longipalpis; espectroscopia de infravermelho

Lutzomyia longipalpis and Lutzomyia cruzi are two phlebotomine species in Brazil that are vectors of Leishmania infantum, the causative agent of visceral leishmaniasis. Lutzomyia longipalpis is found throughout the country, while Lutzomyia cruzi is only present in the Centre-West and Northeast regions. The females of both species are morphologically identical and cannot be differentiated. In Mato Grosso do Sul, both species, Lutzomyia longipalpis and Lutzomyia cruzi, are found in six municipalities where visceral leishmaniasis has been reported. It is important to taxonomically identify the species as the epidemiological profile can be influenced by the vector species. The longipalpis complex, which includes Lutzomyia longipalpis and other species, has been identified based on behavioural, biochemical and morphological evidence. To address this issue, alternative methods must be explored. Fourier Transform Infrared Spectroscopy (FTIR) is a technique that can characterize molecular bands through sound spectra without the need for prior sample preparation. It is commonly used in conjunction with machine learning (ML) by various groups. The objective of this study is to differentiate between female Lutzomyia longipalpis and Lutzomyia cruzi using FTIR and machine learning. A total of 120 female sand flies, 60 Lutzomyia cruzi and 60 Lutzomyia longipalpis, were analysed using a spectrometer in groups of four. The obtained spectra were analysed using multi-analysis and machine learning to characterise the species. The Linear SVM algorithm was used in three band ranges (4000 to 600 cm-¹; 3000 to 2800 cm-¹; 1800 to 800 cm-¹) with the principal components required for each range. The results showed over 95% accuracy. It is also suggested that any spectral range can be used to obtain a good predictive model for use in laboratory routine. The validation tests were successful, with an overall accuracy of 100% for all the spectral ranges analysed with the appropriate choice of PCs. The vibrational bands 2800 cm-¹ (lipids and fatty acids) and 1154, 1109 cm-¹ (carbohydrates) were found to correspond to the differences between the two species. Therefore, it can be concluded that FTIR and machine learning are effective in differentiating the two species.