Novas metodologias espectroscópicas e eletroquímica aplicando classificação multivariada em sistemas biológicos

Abstract

This thesis demonstrates the development of new instrumental methodologies such as: near and medium infrared spectroscopy, molecular fluorescence and differential pulse voltammetry, combined with multivariate classification and variable selection techniques applied to biological system such as tissue, fly larvae and blood plasma. In a first study, NIR spectroscopy was used to obtain healthy and critically defective rat skull spectra. Multivariate classification models were constructed, aiming at an alternative methodology to estimate the success rate of bone repair through low intensity laser and collagen treatments. The algorithms PCA-LDA, SPA-LDA and GA-LDA were applied as classification tools and their comparative performances. In general, the models were able to classify the control group of the others showing efficiency in the evaluation of wound healing in rat cranial defect. In a second study, ATR-FTIR spectroscopy was used to obtain blood plasma spectra of healthy elderly patients with sarcopenia. Multivariate classification models were constructed, aiming at a diagnostic methodology for screening of sarcopenic elderly. The algorithms PCA-LDA, SPA-LDA and GA-LDA were applied as classification tools and their comparative performances. In general, the results obtained through the SPA-LDA and GA-LDA variable selection algorithm were the most satisfactory, achieving sensitivity of 86.7%, since they could be related to functional groups belonging to different biomolecules and possible biomarkers. In a third study, differential pulse voltammetry and molecular fluorescence excitation / emission matrices were obtained for extracts of control fly larvae and containing the drug flunitrazepam. The performance of classification algorithms was evaluated using linear discriminant (90% sensitivity) and quadratic (100% sensitivity) methods for the voltamograms and the 2D-LDA algorithm (sensitivity of 100%) compared to PARAFAC - LDA (sensitivity of 40%) for classification between the groups investigated. The results suggest that spectroscopic and electrochemical methods added to the multivariate analysis are capable of mining relevant information about the understanding of the biological dynamics of several organisms, as well as contributing to the public health system in the identification of sarcopenia in the elderly and classify insects with drugs.