Genetic and phenotypic heterogeneities of psychiatric disorders remain a formidable challenge for understanding circuit mechanisms and accurate diagnosis. Despite tremendous progress in brain imaging techniques, identification of imaging-based biomarkers that helps to guide diagnosis and treatment for neuropsychiatric disorders remains an active topic of research. Towards this end, we develop novel cross-species comparative paradigms that translate genetic, behavioral and neuroimaging findings from diseased monkey models to psychiatric patients. Through this work, we have developed a core set-based machine learning framework that leverages the knowledge gleaned from the genetically-engineered primate models to dissect clinical heterogeneity of psychiatric diseases. We aim to probe the mechanistic links between gene expressions, dysfunction of specific neural networks and symptomatology, thereby identifying a tangible circuit endophenotype that associates aberrant brain connectome with behavioral domains. Such a circuit endophenotype which may cut across multiple categorical diagnoses provides an alternative route to deconstruct inherent heterogeneity and complex comorbidity in mental disorders. In the end, we assess the species difference in this circuit endophenotype through using computational model-based fMRI.