Desenvolvimento de uma plataforma biofotônica acoplada em algoritmos de inteligência artificial para detecção salivar do vírus chikungunya

Abstract

Chikungunya virus (CHIKV) is in the Alphavirus genus of the Togaviridae family. It is an arbovirus whose etiological agent is transmitted through the bite of infected females of the Aedes genus, such as Aedes aegypti and Aedes albopictus. The first documented outbreak of CHIKV occurred in Tanzania between 1952 and 1953, causing a febrile illness accompanied by joint pain. In 2013, CHIKV was introduced to the American continent, triggering a significant epidemic in several Central American countries and the Caribbean islands. In Brazil, the circulation of the virus was first identified in the states of Amapá and Bahia in 2014. Currently, all states in the country report the transmission of this arbovirus. In addition to fluids traditionally used for diagnosis in humans, such as serum and plasma, the presence of CHIKV in other body fluids, including genital secretions, has already been proven. In this sense, saliva emerges as a promising biofluid rich in biological markers, allowing approaches that range from changes in biochemical products to DNA, RNA, and proteins. However, research related to the diagnosis of CHIKV in saliva is still limited. This thesis consists of two chapters that focus on the development of a platform based on ATR FTIR technology, associated with machine learning algorithms, to detect infection caused by CHIKV in saliva samples. In the first chapter, the study aimed to identify specific salivary infrared profile signatures to detect CHIKV infection in a murine animal model, which is suitable for discriminating this emerging disease using multivariate analysis and machine learning algorithms. Thus, interferon-gamma gene knockout C57/BL6 mice were inoculated intradermally with CHIKV (20 µl, 1 x 105 PFU/ml, n=6) or vehicle (20 µl, n=7). Saliva and serum samples were collected on day 3 (due to peak viremia). CHIKV infection was confirmed by real-time PCR in serum from CHIKV-infected mice. The best predictive analysis was using the support vector machine (SVM) algorithm, which presented an accuracy of 85%, sensitivity of 83% and specificity of 86%. In the second chapter, the study aimed to investigate the accuracy of ATR-FTIR to detect CHIKV diluted in different concentrations in human saliva. Therefore, algorithm models were applied to differentiate 6 concentrations of CHIKV in saliva compared to uninfected saliva. Here, the AdaBoost algorithm model showed an accuracy of 83% (sensitivity: 83%, specificity: 83%) for the highest concentration of 4X104 PFU/mL For the second concentration 4X103 PFU/mL, the accuracy was 75% (sensitivity: 66%, specificity: 83%), this concentration being similar to that already reported in a diagnostic study of the virus in saliva. Our results demonstrated that this sustainable biophotonic platform using infrared spectra of saliva, coupled with artificial intelligence algorithms, could detect CHIKV infection in saliva samples.