Tandem 4: Computational Modeling of Disease Mechanisms in Affective and Nonaffective Psychoses

Short Summary:

Using multi-modal neuroimaging, like pharmacological fMRI and EEG, in conjunction with computational models, we study maladaptive cognition in healthy volunteers and psychiatric patients, with a focus on (non-)affective psychoses and impulsivity.


This project concerns mathematical modeling of maladaptive cognition in affective and non-affective psychoses. Using a combination of computational models – e.g., hierarchical Bayesian models (Mathys et al., 2011), dynamic causal models of neuronal population activity (Stephan et al., 2008) and machine learning techniques (Brodersen et al., 2011) – we aim to infer on mechanisms of aberrant cognition in individual patients from neuroimaging data. The hope is that model-based estimates of neuromodulation and synaptic plasticity will enable a dimensional (and physiologically interpretable) classification of distinct disease subgroups in psychosis. In recent proof-of-concept studies (Moran et al., 2011; Schmidt et al., 2013), we have demonstrated the possibility of inferring synaptic processes from “macroscopic” neuroimaging data. We now plan to conduct further pharmacological validation studies which are tailored to putative pathophysiological processes of affective and non-affective psychoses and are complemented by genetics. These studies will exploit multi-modal neuroimaging techniques (fMRI, EEG) at the Translational Neuromodeling Unit (TNU) and the Psychiatric University Hospital (PUK) and will provide a basis for longitudinal studies of patients with affective and non-affective psychosis.

Specific Aims:

(i) Pharmacological and genetic validation studies of computational assays of neuromodulation and synaptic plasticity in healthy volunteers. (ii) Cross-sectional neuroimaging studies of patients with affective and non-affective psychoses. (iii) Verifying the robustness and sensitivity of model predictions in longitudinal neuroimaging studies of patients.

Molecular imaging techniques used:

Pharmacological fMRI and EEG, in conjunction with computational assays.

Conceptual overview of model-based clustering by generative embedding (Brodersen et al. 2014).

Added value of KFSP for this specific tandem project:

To support other projects in the KSFP, we offer: (i) multi-modal neuroimaging facilities (fMRI, EEG, pharmacology, brain stimulation), (ii) computational neuroimaging expertise (incl. a weekly 'methods clinic'), (iii) facilities for translational research with mouse models (genetics, behaviour).