The consumption of poor-quality protein increases the risk of essential amino acid (EAA) deficiency, particularly for lysine and threonine. Thus, it is necessary to be able to detect easily EAA deficiency. Objective: The purpose of this study was to develop metabolomic approaches to identify specific biomarkers for an EAA deficiency, such as lysine and threonine. Three experiments were performed on growing rats. In Experiment 1, rats were fed for 3 weeks with lysine (L30), or threonine (T53) deficient gluten diets, or non-deficient gluten diet (LT100) in comparison with the control diet (milk protein, PLT). In Experiments 2a and 2b, rats were fed at different levels of lysine (L) or threonine (T) deficiency: L/T15, L/T25, L/T40, L/T60, L/T75, P20, L/T100 and L/T170. 24h urines and blood samples from portal vein and vena cava were analyzed using LC-MS. Data from Experiment 1 were analyzed by untargeted metabolomic and Independent Component - Discriminant Analysis (IC-DA) and data from Experiments 2a and 2b by targeted metabolomic and a quantitative PLS regression model. Each metabolite identified as significant by PLS or ICDA was then tested by one-way ANOVA to evaluate the diet effect. Two phases linear regression analysis was used to determine lysine and threonine requirements. ICDA and PLS found molecules that discriminated between the different diets. A common metabolite, the pipecolate, was identified in Experiments 1 and 2a, confirming that it could be specific to lysine deficiency. Another metabolite, taurine, was found in Experiments 1 and 2b, so probably specific to threonine deficiency. Pipecolate or taurine breakpoints obtained gives a value closed to the values obtained by growth indicators. Our results showed that the EAA deficiencies influenced the metabolome. Specific urinary biomarkers identified could be easily applied to detect EAA deficiency and to determine which AA is deficient.