Mapping Cognitive and
Motivational Domains Across Levels
of Analysis:
Challenges and Opportunities
for Target Specification
Robert M Bilder, PhD
Michael E. Tennenbaum Family Professor, and
Chief of Medical Psychology – Neuropsychology,
UCLA Jane & Terry Semel Institute for Neuroscience & Human Behavior,
Stewart & Lynda Resnick Neuropsychiatric Hospital,
Departments of Psychiatry & Biobehavioral Sciences and Psychology
David Geffen School of Medicine at UCLA, and College of Letters & Science at UCLA
Rzhetsky et
al 2009
“genetic
overlap”
derived
from 1.5M
medical
records
Mapping Research Domains
Impact of the NIMH Research Domains Criteria (RDoC) Initiative
• RDoC – offers new dimensions to move
psychiatry beyond current diagnostic taxonomy
and towards rational therapies
• Consensus meetings generated lists of elements
at different levels of analysis
• Challenges: defining the true causal paths that
lead from level to level (genome to syndrome)
• Opportunities: defining targets of treatment
aligned with biological and psychological science
Self-reports
• Personality, symptoms
Behavior
• Rating scales, cognitive assessments
Physiology
• MRI, EEG, MEG, psychophysiology
Circuits
• Neuroanatomic nodes; imaging, basic
Cells
• Cell physiology
Molecules
• Genetics, expression, proteomics
RDoC
Levels of
Analysis
RDoC Matrix Specification:
Expectancy/Reward Prediction Error
RDoC Working Memory Matrix
Unit of Analysis
Genes
Molecules
Cells
Circuits
NRG1/Neuregulin
DISC1
DTNBP1/Dysbindin
BDNF
COMT
DRD2
DAT1
Dopamine
D1
D2
Glutamate
NMDA
AMPA ?
GABA
A?
B?
Pyramidal
Distinct Types of
Inhibitory Neurons
Parvalbumin
Calbindin
Calretinin
Medium Spiny
Neurons (Basal
Ganglia)
Key Circuit: PFCParietal-CingulateDorsal ThalamusDorsal Striatum
DLPFC
VLPFC
Dorsal Striatum
Dorsal Parietal
Inferior Parietal
MD & VA Thalamus (by
virtue of their role in
circuit)
Active
Maintenance
Flexible
Updating
Limited
Capacity
Interference
Control
X
X
X
X
X
X
X
X
X
X
X (gain)
?
X
X
X
X
X
X
?
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
?
Active
Maintenance
Flexible
Updating
Limited
Capacity
Interference Control
N-Back
X
X (?)
X
X (if you include non-target lures)
Delayed Match to
Sample
X
---
X
X (if you use repeated items, or delay
period interference)
Delayed Match to NonSample
X
---
X
X (if you use repeated items, or delay
period interference)
Sequence Encoding and
Reproduction
X
---
X
---
X
X (recent
negative
task
increases
demand on
updating)
X
X (if you use repeated items, recent
negative variation)
X
X
X
X
X
X (?)
X
X (?)
X
X
X
X
X
---
X
X (if you use concurrent interference,
as in Digit Span Distraction)
X
X
X
X
--X
X
X (?)
X
X
X
X
--X
--X
Unit of Analysis
Sternberg Item
Recognition (including
recent negative
variations)
Behavior and Paradigms
Complex Span Tasks
Letter Memory/Running
Memory
Letter Number
Sequencing
Simple Span Tasks (may
be more appropriate for
developmental
populations, in adults
may not capture all key
elements of WM)
Change Detection Tasks
Keep Track Task
AX-CPT/DPX
Self-Ordered Pointing
Is there a better way?
Classic (psychometric) approach
Network (causal modeling) approach
chronic stress
depressed
mood
self-reproach
insomnia
fatigue
concentration
Borsboom & Cramer 2013 Annual Rev Psychology
“In sum, not only do we not know
that symptoms are caused by mental
disorders, but it is in fact extremely
unlikely that they are. As a result, the
hypothesis that such disorders are the
proper entities to steer the
organization of research, diagnosis,
and treatment is, at best, awaiting
scientific justification.”
Borsboom & Cramer 2013 Annual Rev Psychology
What are the
proper entities?
Can we link those
to real behavior in
human1 patients2?
1homo
sapiens
2with real-life problems
RO1MH082795, and Consortium for Neuropsychiatric Phenomics (UL1DE019580, PL1MH083271, RL1LM009833)
Architectures for cognitive ontology
development
The Cognitive Atlas is
conceptualized as a related set
of maps. A given map may
contain sets of related
concepts, quantitative models
of literature association,
annotated effect size statistics,
raw data, summaries of voting,
and qualitative free-text
inputs.
For cognitive concepts (e.g.,
the “phonological buffer”)
there are associated cognitive
concepts, and a “test” layer
comprising objective indicators
of the concepts
RO1MH082795, and Consortium for Neuropsychiatric Phenomics (UL1DE019580, PL1MH083271, RL1LM009833)
It might be argued that the task of the
psychologist, the task of understanding behavior
and reducing the vagaries of human thought to
a mechanical process of cause and effect, is a
more difficult one than that of any other
scientist.
(D. O. Hebb, 1949, p. xi)
Consortium for Neuropsychiatric Phenomics (UL1DE019580, PL1MH083271, RL1LM009833)
Managing assertions about brain-behavior relations using a
neural circuit description framework
Bilder, Howe & Sabb, 2013
Journal of Abnormal Psychology
Multi-Level Assays of Working Memory and Psychopathology: R01 MH101478
Models to Validate Circuit Constructs
Multi-Level Assays of Working Memory and Psychopathology: R01 MH101478
Validating Cognitive Markers of CircuitLevel Constructs
Mapping to Functional Status
Do symptoms or diagnosis add useful prediction over basic
measures of circuit, cognitive or neuropsychological measures?
Diagnosis
(DSM-V)
Symptom
(CIDI, DSM-CC,
BPRS, PROMIS)
Neuropsych
(5 indicators)
Cognitive
(8 indicators)
Circuit
(5 fMRI, 3 sMRI/
DTI, 6 EEG
indicators)
Disability
(WHODAS 2.0)
To avoid extreme group bias,
sampling strategy is agnostic to
diagnosis, and comprises two
groups:
- Care-seeking
- Not Care-seeking
Diagnoses assigned after
enrollment, as one of the
dependent variables under study
Multi-Level Assays of Working Memory and Psychopathology: R01 MH101478
Matching Measurements to Samples
The statistical
power for a given
rating scale may be
reduced if it is
applied in a mixed
population (e.g.,
PANSS in SZ+BP).
A new endpoint
(e.g., ‘g’ from
bifactor model) may
show greater
invariance across
samples, and thus
increase power.
Ariana Anderson, supported by Janssen R&D/UCLA
Pharmacogenomics Research Collaboration
Conclusions
• RDoC “matrix” implies causal links across multiple
levels from genome to syndrome
• Causal models can help specify what to measure
and how to measure
• Validation can proceed directly to function if the
functions and populations are well defined
• New paradigm: specify models of dimensions and
targets without disease entity assumption
• Caveat: new models must be explicit
Many thanks!
[email protected]
http://www.semel.ucla.edu/creativity
http://www.phenomics.ucla.edu
Consortium for
Neuropsychiatric
Phenomics investigators
(52) including PI’s: Freimer,
Cannon, London, Jentsch,
Parker, Evans
Cognitive Atlas
investigators including:
Poldrack, Toga, van Horn,
Sabb
RDoC WM investigators
including: Rissman, Loo,
Bearden, Gitlin, Makeig
Janssen R&D/UCLA team
including: Anderson,
Salvadore, Chung, Wilcox,
Savitz, Alphs, Wang, Li