Skip to main content

Transforming the understanding
and treatment of mental illnesses.

Celebrating 75 Years! Learn More >>

Applying Computationally-Defined Behaviors for Back-Translation in Psychiatry


Michele Ferrante, Ph.D.; Division of Translational Research and Division of Neuroscience and Basic Behavioral Science
Janine Simmons, M.D., Ph.D.; Division of Neuroscience and Basic Behavioral Science


To apply computational approaches to develop parametrically detailed and back-translatable behavioral assays across mental health relevant domains of function.


Defining the neurobiological processes underlying the regulation and dysregulation of human behavior requires the ability to investigate questions critical to mental health in both humans and animals. Ideally, tools available for psychiatric and behavioral neuroscience research would include a library of behavioral assays that can be used in both humans and rodents to assess mental health relevant domains of function and to test hypotheses regarding neurobiological mechanisms. Currently, significant gaps exist in our ability to move research bi-directionally between humans and rodents.

Behavioral and cognitive assays designed for screening medications in rodents (e.g., forced swim test, tail-suspension test, sucrose preference) are rarely predictive of human outcomes. Complex paradigms or self-report measures designed for use in humans present obvious barriers for back-translating to rodents. Moreover, current paradigms for measuring human behavior often lack the computational rigor necessary to reliably model the richness and variability critical to studies of mental health and illness. These problems hinder our translational pipeline.

To address this gap, NIMH is interested in the development and deployment of a new set of computationally-informed behavioral paradigms. Such paradigms would be based upon theoretical frameworks and mathematical models able to account for quantitative, parametric behavioral measurements. These assays need to be developed in humans first, keeping in mind applications of the same models in rodents, to strengthen our translational pipeline. The use of computationally-informed paradigms would allow behaviors to be captured, segmented, and classified in a highly parameterized fashion. These systems would provide high-level objectivity and consistency for a better understanding of behavioral dynamics and more rigorous hypotheses generation and testing of neurobiological mechanisms.

Projects responsive to this concept would include:

  • A well-defined question in behavioral science relevant to psychiatric populations.
  • Behavioral targets, models, and parameters that reflect a dimensional process linked to a specific domain of function and have the potential for back-translation from humans to rodents.
  • Complex behavioral targets expressed as the outcome of specific computations through the application of rigorous mathematical modeling i.e., formulating mathematically and experimentally tractable behavioral questions.
  • Highly-parameterizable, mental-health relevant behavioral targets that readily lend themselves to computational analysis, predictions, and explanations and that have not been extensively mathematically modeled (e.g. learning theory).
  • A theoretical model of a behavioral phenotype that can be achieved by:
    • Breaking down the behavior in fine-grained parameters that can be mathematically described.
    • Integrating the behavioral parameters in an experimentally-grounded mathematical formalism (e.g., a new behavioral theory). The model should integrate previous experimental findings, allow the tracking and integration of all parameters, and predict behavioral outcomes over time.
    • Experimentally validating and optimizing these theory-driven models. The mathematical model needs to be experimentally tested, validated, and refined by rigorous determination of the relationships between all the behavioral parameters and the outcome variables. This process includes demonstration that the model can make behaviorally accurate predictions for a new experiment or dataset.