Application of the Gradient Descent Method for Optimization of Spin System Parameters of Metabolite Molecules by NMR Spectra

Nuclear Magnetic Resonance Spectroscopy is an effective physical method for metabolite fingerprinting, which involves the simultaneous and extensive analysis of a wide variety of compounds. The main disadvantage of the method, which hinders the progress of NMR metabolomics, is the need for manual processing of complex NMR spectra of biological samples. To automate the identification of metabolites in spectra, it is necessary to determine the parameters of the spin system of the main metabolites found in the samples under study. To address this issue, we propose an optimization algorithm that autonomously optimizes all relevant spin system parameters. The algorithm’s successful operation has been demonstrated on molecules of the most common amino acids found in NMR spectra of biological samples. The developed algorithm produced a proline spin-spin coupling matrix that, when evaluated by the Intersection-over-Union metric, showed better consistency with five experimental NMR spectra than literature matrices.