NMR Phase Error Correction with New Modelling Approaches

Abstract

Nuclear Magnetic Resonance (NMR) spectroscopy is a highly sensitive analytical technique essential for precise molecular identification and quantification. However, accurate results depend on effective pre-processing to correct for various types of errors. Phase error correction, in particular, is crucial for ensuring the reliability of NMR data. Current methods often rely on a single linear model, which may not adequately address all types of phase errors. As a result, this limitation frequently requires manual intervention, making the process both time-consuming and prone to errors. To address these limitations, we propose three modelling approaches for NMR phase error correction: nonlinear shrinkage, multiple models, and a new optimization function called delta absolute net minimization (DANM). Our comparison of seven methods revealed that nonlinear shrinkage outperformed others in both simulated spectra and a diabetes study, followed by multiple models with DANM. Additionally, our spike-in experiments demonstrated that DANM performed quite well in both single and multiple models. Our nonlinear shrinkage approach is a simple yet effective solution. We provide an open-source R package, NMRphasing, available on CRAN (https://cran.r-project.org/web/packages/NMRphasing/) and on GitHub (https://github.com/ajiangsfu/NMRphasing).