Peak alignment of one-dimensional NMR spectra by means of an intensity fluctuation frequency difference (IFFD) segment-wise algorithm

Abstract

The increasing scientific and industrial interest in metabolomics often takes advantage of the high level of qualitative and quantitative information provided by nuclear magnetic resonance (NMR) spectroscopy. However, several chemical and physical factors can affect the frequency of an NMR resonance. Especially in complex biological samples such as biofluids, small perturbations in NMR chemical shifts can complicate the recovery of biomarker information in metabolomics studies using multivariate statistics and pattern recognition tools. Novel segment-wise peak alignment algorithms have been proposed in the literature to correct the misalignment of NMR signals. The approach presented here, the Intensity Fluctuation Frequency Difference (IFFD) algorithm, is a highly efficient method designed to reduce variability in peak positions across the multiple NMR spectra used in metabolomics studies. This automated method refines segmentation using differences in the frequencies of the intensity fluctuations of signals and baseline noise to improve spectral alignment. Alignment is performed sequentially using an open source program, icoshift, employing a fast Fourier transform (FFT) engine to align all spectra simultaneously. The IFFD-icoshift method is illustrated for ¹H NMR spectra measured for 50 human urine samples collected from healthy volunteers: 41 samples, including urine from a pregnant female, were collected randomly following a normal dietary routine and 9 samples were collected after dietary supplementation with ibuprofen, alcoholic beverages or an energy drink. We demonstrate the superior performance of IFFD-icoshift alignment over a wide range of peaks and its capacity to enhance the interpretability and robustness of multivariate statistical analysis. This approach is widely applicable for NMR-based metabolic studies and is potentially suitable for many other types of data sets such as chromatographic profiles and MS data.