A chromatographic retention assisted denoising and peak picking algorithm (CRAD) is developed for preprocessing liquid chromatography-mass spectrometry (LC-MS) datasets of natural products. The retention behaviors of ions with the same m/z value are investigated under a series of elution conditions. The detected ions are identified as real compounds if their chromatographic retention behaviors fit well with the Snyder-Soczewinski model. Further, the ions with similar retention behaviors and isotope ratios are clustered. This method enables rapid identification of precursor ions when chemical standards or databases are unavailable. It also helps eliminate unexpected baseline disturbances and improve the resolution of LC-MS chromatograms. Unlike conventional deconvolution strategies, this method distinguishes the chemical properties of precursor ions through their dynamic retention behaviors. The algorithm is demonstrated with LC-MS datasets of control samples. In the application of such algorithms on a more complicated natural extract from Lycium ruthenicum Murr., 206 precursor ions were facilely determined.