From Interoceptive Insufficiency to Computational Cures: Modeling, Mapping, and Modulating Brain Circuits in Affective and Substance-Use Disorders
Neuroscience Seminar Series

Abstract:
There are profound challenges for psychiatric nosology by transdiagnostic symptoms, such as anhedonia, anxiety, and impaired decision-making. These symptoms cut across affective, anxiety, and substance-use disorders, suggesting common underlying neural substrates that are not adequately captured by current diagnostic frameworks. We propose a conceptual model that bridges this gap by integrating three distinct methodological domains: (1) Neurobiological Mapping, (2) Computational Phenotyping, and (3) Targeted Neuromodulation. Several empirical studies will be presented providing evidence implicating a core "interoceptive-decision-making axis"—centered on the insular and anterior cingulate cortices—in processing internal bodily states and biasing choices, particularly under uncertainty. Our group has shown how this mapped circuit-level dysfunction can be deconstructed into precise computational parameters using machine learning and advanced statistical approaches, and identifying transdiagnostic contributors to symptoms like repetitive negative thinking. Finally, emerging evidence will be presented that these computationally-defined circuit deficits can be directly and causally targeted using novel, non-invasive neuromodulatory interventions, such as low-intensity focused ultrasound (LIFU) and repetitive transcranial magnetic stimulation (rTMS), to ameliorate symptoms in substance-use and affective disorders. These findings will be used to propose a mechanistic model where computational psychiatry provides the formal bridge from neural-systems mapping to principled, circuit-based therapeutic intervention, offering concrete, testable predictions for the next generation of translational biomarkers.

About the Speaker:
Dr Martin Paulus is the Scientific Director of the Laureate Institute for Brain Research (LIBR), where he leads a translational neuroscience program integrating clinical phenotyping, neuroimaging, and biomarker science with real-world clinical data. At LIBR, he is advancing the application of AI in academic psychiatry—using machine learning, large language models, and causal inference methods to extract clinically actionable insights from large-scale electronic health records and longitudinal patient-reported outcomes. His work emphasizes rigorous validation, interpretability, and deployment-relevant design so that AI models can meaningfully inform treatment selection, risk stratification, and precision psychiatry workflows. On this panel, he will discuss practical strategies for building reproducible AI-enabled research pipelines in academic settings and translating them into measurable gains in patient care and discovery.

Contact

  Stephen Wilson
  sjw42@psu.edu