EDMONTON, Alberta—As analytical tools improve, the use of metabolic information in drug discovery and development increases. Challenges in NMR peak identification in complex samples limits, however, the widespread application of the method in the lab. To address this problem, researchers at Chenomx and the University of Calgary recently developed a software solution that relies on correlations of experimental and idealized NMR spectra to identify individual metabolites. They presented their work in Analytical Chemistry.
To improve the accuracy of the statistical pattern recognition tools, the researchers replaced the traditional spectral binning operation—whereby complex NMR spectra are broken into "bite-sized" chunks that can be more easily analyzed—with a method they call "targeted profiling". The new method relies on producing computational models of metabolite spectra that are then combined to form a cumulative model spectrum for comparison with the actual sample spectrum. They then compared the two methods on a sample comprised of various metabolites at physiological concentrations, designed to mimic a typical biofluid.
Although targeted profiling was time-consuming—the method has yet to be automated—it compared favorably to spectral binning and the researchers noted that the new method was superior when examining the overlapped spectra of low-concentration metabolites. For this reason, the researchers don't see targeted profiling as a replacement for spectral binning, but rather as a complementary method that should improve metabolic profiling.