Cluster-Based 1H NMR Alignment with Joint Phase-Baseline Optimization: A Physically Constrained Pipeline for Metabolomics and Food Authentication

Abstract

Accurate alignment of one-dimensional 1H NMR spectra is a prerequisite for reliable metabolomics, but chemical shift variability, line shape asymmetry, and multiplet overlap continue to compromise conventional warping algorithms. In this study, we propose a fully automated, cluster-based alignment framework that enforces the physical constraints of scalar coupling. Within a single optimization loop, zero- and first-order phase parameters are refined, while the baseline is dynamically re-estimated. Peaks are extracted with a matched filter derived from an in-spectrum singlet, and multiplets are recognized by a jump-detection criterion applied to a cluster-distance vector. Optimal peak-to-peak correspondence is then established under coupling-constant, binomial-intensity and coherent chemical-shift-variability rules, and a shift-corrected spectrum is reconstructed by cubic-spline interpolation. Validation on simulated spectra exhibiting severe “crossing chemical-shift variability” demonstrates accurate recovery of multiplet patterns. When applied to 81 1H NMR spectra of Lycium barbarum L. from three geographical origins, PCA demonstrates improved alignment accuracy and enhanced variance interpretation and geographical group discrimination. It is implemented in open-source Python for vendor format import, with demonstrated applicability to metabolomics or food-quality workflows.