Traditionally, symptom-based classifications define mental health diagnoses. However, this approach can oversimplify clinical presentations, and these diagnoses often do not correspond with biological markers or meaningful clinical outcomes. A growing body of literature seeks to identify neurobiological subtypes (e.g., based on neurostructural features) to reflect biological representations of psychopathology. Even still, this research often demonstrates a reliance on traditional diagnoses in order to define psychopathology. To address these limitations, the present studies utilize a machine learning approach, heterogeneity through discriminative analysis, coupled with a dimensional model of psychopathology to identify neurobiological subtypes that embrace the continuous, transdiagnostic, and hierarchical nature of psychological symptoms. Using a large sample of 9-to-10-year-old children (n = 9,027) followed across two years, these studies delineate neurostructural subtypes among children who highly endorse broad psychological concerns (Study 1) and use deep phenotyping to explore distinctions among the subtypes’ neural characteristics, cognitive ability, and associated psychopathology traits both cross-sectionally and longitudinally (Study 2). Taken together, this data-driven approach reveals neural heterogeneity as demonstrated by structural patterns that map onto divergent profiles of psychopathology symptoms and cognitive performance in youth. This phenotyping of psychological disorders may aid the identification of potential neurobiological mechanisms and clinical presentations which could ultimately improve diagnostic accuracy and treatment outcomes.