Individually Measured Phenotypes to Advance Computational Translation (IMPACT)

Presenter:

Jenni Pacheco, Ph.D.
Research Domain Criteria (RDoC) Unit and Division of Translational Research

Goal:

This concept aims to encourage research using novel behavioral measures to foster a new generation of clinical signatures, leading toward precision assessment, prognosis, and treatment of mental disorders. The initial focus will be to (1) develop or optimize behavioral tasks that measure individual differences and demonstrate added utility for clinical prediction when combined with standard clinical diagnosis; and (2) form a data infrastructure that can support computational approaches to build tools for clinical decision making. This research will involve one or more longitudinal cohorts established either through new data collection or by leveraging extant cohorts that have appropriate data structures.

Rationale:

This initiative is intended to enable transdiagnostic, data-driven precision assessments in psychiatry. It will support studies that follow large cohorts of individuals with a range of psychopathology over time using innovative behavioral tasks, multimodal measures, and computational methods to refine clinical assessments, improve prognosis, and optimize treatment in persons with mental disorders. Data-centric and data-driven approaches can be used to generate more specific and less biased clinical phenotypes, which can be subtypes of individual disorders or transdiagnostic, cutting across two or more traditional disorders. The initiative will promote the use of artificial intelligence and machine learning (AI/ML) techniques to identify patterns that are associated with precision diagnoses, clinical outcomes, or transdiagnostic dimensions. Importantly, these computational techniques can help to uncover which measures incrementally add predictive power or diagnostic refinement, paving the way to tailored assessment tools that promote the most accessible and clinically informative measures. From a treatment perspective, these data-driven clinical phenotypes may also help to select more effective and targeted treatment options.

This initiative seeks to address two fundamental needs that will serve as a foundation for identifying meaningful clinical signatures: (1) cultivating a new generation of behavioral tasks that are predictive on the individual level and (2) establishing a data infrastructure that enables computational exploration of multimodal data. Innovative, theory-driven behavioral tasks are needed to offer objective, accessible metrics to identify phenotypes related to functional impairments. NIMH aims to focus on enhancing predictive power at the individual-subject level, the capacity for robust remote administration, and a demonstrable relationship to an underlying neural circuit or network. In addition, to understand the predictive nature of these phenotypes, a large-scale data infrastructure is needed that includes longitudinal data from multimodal measurements, combined with a record of clinical treatments and outcomes. These different types of data will be used to detect patterns of clinical trajectories, with the eventual goal of developing quantitative tools to be tested by clinicians in making decisions for individual patients. The objectives of this initiative are consistent with the NIMH RDoC initiative and align with several key NIMH priorities and Strategic Goals.