Human blood lipoprotein predictions from 1H NMR spectra: protocol, model performances and cage of covariance

Abstract

Lipoprotein subfractions are biomarkers for early diagnosis of cardiovascular diseases. The reference method, ultracentrifugation, for measuring lipoproteins is time consuming and there is a need to develop a rapid method for cohort screenings. Here we present partial least squares regression models developed using 1H-NMR spectra and concentrations of lipoproteins as measured by ultracentrifugation on 316 healthy Danes. Different regions of the 1H-NMR spectra representing signals of the lipoproteins and different lipid species were investigated to develop parsimonious, reliable and best performing prediction models. 65 LP main and subfractions were predictable with an accuracy Q2 of > 0.6 on test set samples. The models were tested on an independent cohort of 290 healthy Swedes with predicted and reference values matching by up to 85-95%. The software was developed to predict lipoproteins in human blood using 1H-NMR spectra and made freely available to be applied for future cohort screenings.