Metabolômica e inteligência artificial no estudo da infecção por Crypto-coccus spp

Abstract

Cryptococcosis affects mainly immunosuppressed hosts, of which HIV-infected individuals represent the majority of cases worldwide. Cryptococcal meningitis (CM) is the most common clinical menifestation in these individuals and evolves with high morbidity and mortality rates, so there is a great interest in the search for new methods for early diagnosis, which can improve the outcome and prognosis of patients. The aim of this study was to develop new diagnostic strategies for cryptococcosis using artificial intelligence, mass spectrometry and biosensors. The metabolomics approach included serum and/or cerebrospinal fluid (CSF) samples from 82 individuals distributed in HIV-infected with and without CM (HIV+CM+(28) and HIV+CM-(45)) and non-HIV without CM (HIV-CM-(9)) who had different neurological conditions. The metabolites were extracted and analyzed by High Performance Liquid Chromatography hyphenated with Mass Spectrometry. The normalized intensities of significant metabolites were compared among the three groups. It was developed an algorithm in which CSF in positive ionization mode in Random Forest was the best predictor to differentiate the three groups, with 93.86% of accuracy. Eight compounds expressed significant differences among the three groups in CSF. The expression of the compound (2-Amino-3-methylphenyl)(4-((4-((3-chlorophenyl)sulfonyl)phenyl)methyl)(1,4'-bipiperidin)-1'-yl)methanone was higher in the HIV+CM- than the HIV+CM+ group. According to the statistical analysis, 41 compounds were considered significant in terms of expression in serum samples among the three groups. In general, the expression of metabolites was higher and more homogeneous in the HIV-CM- group, for both samples. Were identified a total of 43 metabolites, whose molecular weight ranged from 214.26 to 1060.2 g/mol. These compounds belong to different chemical classes, of which lipids were predominant. Data related to metabolomics and cryptococcosis are scarce and to our knowledgement, the compounds here found have not yet been described in this infection. Lipids have natural biological functions in several metabolic pathways and as the field of lipidomics advances, the mechanisms of how these biomolecules affect pathophysiological processes become more clear. Further in-depth studies are needed to confirm these findings, as well as to validate their potential as biomarkers for the diagnosis of cryptococcosis. In addition, Phage Display technique was performed aiming to obtain peptides binding to GXM, the main capsular polysaccharide antigen of Cryptococcus spp. The selected phages were titered, amplified, purified and had their DNA sequenced. There were no frequent motifs among the sequences of the peptides obtained, nor similarity with proteins deposited in available databases. The selected phages were submitted to immunoassay to evaluate the vii reactivity against GXM. The clone with the best performance (H3) was chosen for the continuity of the tests, however, when tested on biological samples it did not show reliable results, a fact that hindered to built a biosensor for the diagnosis of the infection.