Recursive Segment-Wise Peak Alignment of Biological ¹H NMR Spectra for Improved Metabolic Biomarker Recovery

Chemical shift variation in small-molecule (1)H NMR signals of biofluids complicates biomarker information recovery in metabonomic studies when using multivariate statistical and pattern recognition tools. Current peak realignment methods are generally time-consuming or align major peaks at the expense of minor peak shift accuracy. We present a novel recursive segment-wise peak alignment (RSPA) method to reduce variability in peak positions across the multiple (1)H NMR spectra used in metabonomic studies. The method refines a segmentation of reference and test spectra in a top-down fashion, sequentially subdividing the initial larger segments, as required, to improve the local spectral alignment. We also describe a general procedure that allows robust comparison of realignment quality of various available methods for a range of peak intensities. The RSPA method is illustrated with respect to 140 (1)H NMR rat urine spectra from a caloric restriction study and is compared with several other widely used peak alignment methods. We demonstrate the superior performance of the RSPA alignment over a wide range of peaks and its capacity to enhance interpretability and robustness of multivariate statistical tools. The 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.