1/5-Cognitive Neurocomputational Task Reliability & Clinical Applications Consortium

Advancements in computational psychiatry allow us to isolate multiple, specific cognitive mechanisms that
determine human behavior. This formal modeling framework generates quantitative parameter estimates that
can serve as bridges between pathophysiology and psychopathology. A major goal of computational psychiatry
is to translate these laboratory tools so that they can be used in the clinic. Two critical hurdles need to be
overcome. First, the enhanced validity and sensitivity of computational metrics needs to be established relative
to standard behavioral performance metrics in key psychiatric and nonpsychiatric populations. We propose to
do that by addressing a range of cognitive and motivational domains that have been strongly implicated in
psychopathology, including working and episodic memory, visual perception, reinforcement learning, and effort
based decision making. Second, we need to establish and optimize the psychometrics of these computational
metrics so that they can be used as tools in treatment development, treatment evaluation, longitudinal, and
genetic studies. These powerful metrics must have adequate test-retest reliability, and not be limited by ceiling
and floor effects. We propose to develop these methods using an open, flexible, and scalable framework and
demonstrate that they provide valid data both in the laboratory and in large-scale Internet-based data collection,
facilitating “big data” studies of cognitive processes. To this end, the current project will leverage the expertise
of Cognitive Neuroscience Task Reliability and Clinical applications in Serious mental illness (CNTRACS)
consortium, a multi-site research group with an established record of rapid cognitive tool development and
dissemination. Aim 1 is to establish that model based parameters for the measurement of cognitive function are
more sensitive than standard behavioral methods in assessing deficits across a range of common mental
disorders, and have an enhanced capacity to predict clinical symptoms and real-world functioning, with a sample
of 180 patients with psychotic and affective disorders (both medicated and unmedicated) and 100 healthy
controls. Aim 2 is to measure and optimize the psychometric properties (test re-test reliability, internal validity,
floor and absence of ceiling and practice effects) of computational parameters described in Aim 1, in a new
sample of 180 psychiatric patients and 100 healthy controls. Aim 3 is to establish the feasibility and replicability
of model-based analytic approaches outside the laboratory for assessing RDoC dimensions of interest, and to
assess their relationships to variation in psychotic-like experience, depression and anhedonia, as well as real-
world functioning in a community sample of 10,000 recruited over the Internet. Aim 4 is to validate key model
based parameters against well-characterized neurophysiological measures acquired using EEG recordings
during task performance. Successful completion of these Aims will significantly advance the field by providing
easily administered and scalable web-based tools for estimating the integrity of key neural systems that underlie
normal cognition and motivation and form the basis of common forms of cognitive and affective psychopathology.

Grant Number: 5R01MH084840-12
NIH Institute/Center: NIH
Principal Investigator: Deanna Barch