Peak alignment is an important step during metabolomics data pretreatment process based on nucle. ar magnetic resonance (NMR) and its effect plays a direct role on subsequent multivariate statistical analysis. A peak alignment algorithm based on Gaussian smoothing (GPA) is presented. Spectrum signals can be smoothed on multiple scales by adjusting sizes of the windows of Gaussian convolution function. And peak alignment can be re. alized step by step from coarse to fine. The true experiment results of NMR spectrum show that peak alignment can be realized quickly and accurately by GPA algorithm. Comparing with common peak alignment algorithms such as correlation optimized warping (COW) and multi-scale peak alignment (MSPA), the aligned spectrum signals are su. perior at integrated performances such as average similarity and explanation performances of subsequent statistical models obviously.