An analytical strategy for the metabolic profiling of rice grain was developed based on gas chromatography-mass spectrometry (GC-MS). For the purpose of obtaining abundant metabolite information, sample preparation step prior to instrumental analysis is necessary to be optimized. D-optimal experimental design was applied to optimize the extraction solvent. Four solvents, including water, methanol, isopropanol and acetonitrile, and their combinations were evaluated for the extraction efficiency using multivariate statistical analysis (partial least square regression). The count of resolved peaks and the sum of peak areas were taken as the evaluation indexes. Methanol/water (80:20, v/v) mixture was highly efficient for rice metabolites and was selected as the suitable solvent formulation. Then, the analytical characteristics of the method were measured. More than 90% of the metabolites had satisfactory precisions, reproducibilities and stabilities (relative standard deviations (RSDs) < 30%). Most of the detected metabolites (about 88.0% of total peak area) showed good linear responses. With the optimized analytical protocol, 315 metabolites were detected in rice and 86 of which were structurally identified by searching in the NIST 08/Wiley standard mass spectral library, covering carbohydrates, amino acids, organic acids, steroids and so on which showed a broad coverage of metabolite data. The established method is expected to be useful for the metabolomic studies of rice.