Chemometric strategy for aligning chemical shifts in 1H NMR to improve geographical origin discrimination: A case study for Chinese Goji honey

Abstract

1H nuclear magnetic resonance (NMR)-based untargeted metabolomics has been extensively used for the geographical discrimination of food samples. In analyzing complex samples, this technique faces the challenges of baseline drift and NMR peak position shifts, resulting in inaccurate geographical discrimination by current chemometric models, like partial least squares. To address this problem, we provide a novel automatic untargeted chemometrics strategy named AntDAS-NMR in this work. AntDAS-NMR utilizes the originally acquired NMR spectra as inputs to automatically perform baseline drift correction, NMR peak detection, NMR peak-based spectrum alignment, and geographical discrimination analysis. The developed strategy is used for the geographical discrimination of Goji honey samples from the northwest zones of China. Results indicate that with the aid of AntDAS-NMR, the artifacts of baseline drifts and NMR peak shift that influence the geographical discrimination model can be reasonably resolved and that Goji honey samples from various provinces of China can be satisfactorily classified. Additionally, the performance of AntDAS-NMR is comparable with classic NMR data analysis tools, like COW, icoshift, PAFFT, and RAFFT. In conclusion, the AntDAS-NMR may be used as a candidate for 1H NMR-based untargeted metabolomics.