Design of lipid prediction PLS models based on 1H-NMR spectroscopy of human serum samples

Identifier:  TFG:5715

Handle: https://hdl.handle.net/20.500.11797/TFG5715

Authors:  Santiago Gómez Coso

Abstract:
Cardiovascular diseases (CVD) currently have a high influence on the world's population. Advanced analytical technologies such as nuclear magnetic resonance (NMR) have been applied to predict standard lipids, which are strongly related to cardiovascular risk. The current study shows the development and evaluation of 1H-NMR-based partial least squares regression (PLS) models to predict 4 standard lipid variables in human serum samples. The regions with the highest correlation with the concentrations of each lipid variable were selected and then, a training set of samples were used to train the models, while the remaining served to validate them. In addition, a z-score standardisation was performed to decrease biological variability between samples, which led to an improve of the estimation models.