Comparative 1H NMR-based metabonomic analysis of HIV-1 sera

(1)H NMR spectroscopy of sera from HIV-1 infected and uninfected individuals was performed on 300 and 600 MHz instruments. The resultant spectra were automatically data reduced to 90 and 180 integral segments of equal length. Analysis of variance identified significant differences between the sample groups, especially for the samples analyzed on 600 MHz and reduced to fewer segments. Linear discriminant analysis correctly classified 100% of the samples analyzed on the 300 MHz NMR (reduced to 180 segments); an increase in instrument sensitivity resulted in lower percentages of correctly classified samples. Multinomial logistic regression (MLR) resulted in 100% correct classification of all samples from both instruments. Thus (1)H-NMR metabonomics on either instrument distinguishes HIV-positive individuals using or not using anti retroviral therapy, but the sensitivity of the instrument impacts on data reduction. Furthermore, MLR is a novel multivariate statistical technique for improved classification of biological data analyzed in NMR.