To date, no molecular biomarker exists for any psychiatric disorder. To identify phenotype-specific biomarkers and investigate the molecular underpinnings of anxiety pathophysiology, we interrogated the well-established mouse model of high (HAB), normal (NAB) and low (LAB) anxiety-related behavior by in vivo (15)N metabolic labeling and quantitative proteomics. The (15)N metabolic labeling approach enables accurate quantification due to the early mixing of the labeled and unlabeled samples under comparison, thus avoiding the biased experimental error introduction during handling. Differentially expressed proteins between HAB and LAB mice can be validated with non-mass-spectrometry-based methods. In silico pathway analysis enables identification of protein networks implicated in anxiety neural circuits. The presented workflow provides a precise and non-hypothesis-driven tool for identifying candidate biomarkers using animal models of psychiatric disorders.