Self-Modeling Curve Resolution Recovery of Temporal Metabolite Signal Modulation in NMR Spectroscopic Data Sets: Application to a Life-Long Caloric Restriction Study in Dogs

摘要

A novel model-free statistical approach (self modeling curve resolution, SMCR) has been applied to recover biochemical information from complex overlapping signals in ~(1)H NMR spectra of blood serum in a long-term study of caloric restriction (CR) in the dog (n velence 24 control fed (CF) and n velence 24 CR animals). A new statistical spectroscopic construct, the spectrotype, is proposed which is a spectroscopic subset description or component of a metabolic phenotype. Characterization of the ~(1)H NMR profiles according to their evolutionary contribution of each spectrotype gives clues to the kinetics of the macro-biochemical response profiles and the identity of the underlying biochemical constituents, governing the evolutionary global response to an intervention. This information can be used to monitor and predict the end point of the biological process and to identify the mechanisms responsible for those changes. Here a SMCR strategy together with a pattern recognition method, principal component analysis (PCA) was used to resolve sets of spectrotypes, without a priori information. From the ~(1)H NMR evolutionary response profiles, two spectrotypes were identified and resolved; spectrotype 1 dominated by lipids featuring contributions from phosphatidylcholine, lipoprotein lipid fatty acyl groups from triglycerides, phospholipids, and cholesteryl esters plus total cholesterol (i.e., both esterified and unesterified); spectrotype 2 comprising glucose signals and a poorly resolved envelope of albumin and N-acetylated glycoprotein resonances. The relative contributions of these spectrotypes in each sample were calculated. For both caloric restricted (CR) and control fed (CF) dogs between ages 1 and 9 years, the contribution of spectrotype 2 > spectrotype 1, whereas for dogs aged between 9 and 12 years spectrotype 1 > spectrotype 2. Therefore, SMCR analysis pinpointed ages where nutrition and aging metabolic changes became significant within serum samples as well as providing the individual longitudinal contribution profiles associated with each spectrotype, which could potentially be used as part of a strategy to monitor and predict longevity and morbidity in populations. Hence SMCR is a useful addition to the chemometric "toolbox" for metabolic analysis and should have diverse applications within other biomedical conditions characterized by subtle time-dependent changes.