Phenolamides constitute a diverse class of secondary metabolites that are found ubiquitously in plants and have been implicated to play an important role in a wide range of biological processes, such as plant development and defense. However, spatiotemporal accumulation patterns of phenolamides in rice, one of the most important crops, are not available, and no gene responsible for phenolamide biosynthesis has been identified in this species. In this study, we report the comprehensive metabolic profiling and natural variation analysis of phenolamides in a collection of rice germplasm using a liquid chromatography–mass spectrometry-based targeted metabolomics method. Spatiotemporal controlled accumulations were observed for most phenolamides, together with their differential accumulations between the two major subspecies of rice. Further metabolic genome-wide association study (mGWAS) in rice leaf and in vivo metabolic analysis of the transgenic plants identified Os12g27220 and Os12g27254 as two spermidine hydroxycinnamoyl transferases that might underlie the natural variation of levels of spermidine conjugates in rice. Our work demonstrates that gene-to-metabolite analysis by mGWAS provides a useful tool for functional gene identification and omics-based crop genetic improvement.