Quantifying Hidden Metabolites in 31P NMR Spectra by Physics-Based Spectral Subtraction

Abstract NMR spectroscopy enables reliable quantification of the metabolites present in a biological sample. The number of metabolites that can be assessed from NMR spectra is often limited by the intrinsically high complexity of the spectra, arising from the large number of molecules paired with their complex spectral patterns. 31P NMR based metabolomics embraces a reduced yet broad view of the metabolism and significantly facilitates the interpretation of the spectra due to a lower number of signals and a larger chemical shift dispersion. However, full quantification of metabolites can still be hindered by the presence of crowded regions. Here we exploit the fact that NMR lineshapes can be simulated precisely from quantum mechanical calculations for disentangling these crowded regions. Iterative subtraction of accurately simulated and fitted spectra of metabolites presenting isolated signals from the experimental spectrum yields meaningful residuals, enabling assignment and quantification of completely hidden metabolites. This approach ended at a noise-level residual spectrum, granting the full quantification of the 31P NMR spectrum. We implemented this approach in MR-SHIFT, a program developed in this work for spin Hamiltonian simulations and spectral fitting, which was then applied to the fully automated analysis of 31P NMR spectra of mouse tissues.