Computational approach to the schizophrenia: disconnection syndrome and dynamical pharmacology

Transcription

1 Computational approach to the schizophrenia: disconnection syndrome and dynamical pharmacology Péter Érdi1,2, Brad Flaugher 1, Trevor Jones 1, Balázs Ujfalussy 2, László Zalányi 2 and Vaibhav Diwadkar 3 1 Center for Complex Systems Studies, Kalamazoo College, Kalamazoo, Michigan 2 Dept. Biophysics, KFKI Res. Inst. Part. Nucl. Phys. Hung. Acad. Sci. Budapest, Hungary 3 Dept. Psychiatry and Behavioral Neurosciences, Wayne State Univ. School of Medicine, Detroit, MI, USA.

2 Content 1. General framework: dynamical diseases 2. Long-term plan 3. Basic behavioral data 4. Basic fmri data 5. fmri data processing 6. A neural model of normal and pathological associative learning 7. Much left to be done

3

4 General framework: dynamical diseases The theory of dynamical diseases emerged from chaos theory Dynamical disease occurs due to the impairment of the control system: associated to abnormal dynamics Develop realistic mathematical models and study effects of parameter changes Neurobiological interpretation Integration of molecular, cellular and system neuroscience Therapeutic strategies

5 General framework: dynamical diseases Alzheimer disease Migraine Normal Pathological Schizophrenia storage and recall of memory traces PFIZER Pharma GmbH 2005 E E State changes in attractor structure pathological attractors fixed point attractor State Parkinson disease periodic attractor Dynamical Dahlem and Chronicle, Neurobiology, 2004 diseases Anxiety Control Epilepsy 3 20? ADHD 1 mv 3 sec. Power Power Frequency Reboxetine (0.3 mg/kg, IV) Frequency Hz Hz Events (Hz) Events (Hz) Time seconds Time seconds I. Aradi, P. Érdi: Computational neuropharmacology: dynamical approaches in drug discovery. Trends in Pharmacological Sciences 27(5) (2006)

6 Long term plan Some schizophrenic phenomena as dysconnection syndrome? Combined behavioral and fmri data Neural model of intereacting regions for normal and impaired learning fmri data analysis: uncover the normal and pathological information flow Computational pharmacology: to shift the system from dynamical pathological state to normal one

7 Basic Behavioral Data Associative Learning ENC RET ENC RET ENC

8 Basic Behavioral Data Associative Learning ENC RET ENC RET ENC

9 Basic Behavioral Data

10 Brain areas involved

11 Basic fmri data fmri Data Encoding (HC) PP V1 ITp MTL MTL ITp Group random effects analysis showing regions of greater average activation during the encoding of objects and locations relative to a fixation baseline (p <.005, uncorrected). Network consists of regions in the Superior Parietal and ITp, primary visual cortex (V1), and the hippocampus (MTL).

12 Basic fmri data fmri Data Schizophrenia Encoding Retrieval

13 fmri data processing Correlation with sensory signal Correlation Brain region red: recall black: learning filled: schizo open: HC Occ SP IT Hpc PFC Fro Cing

14 A neural model of normal and pathological associative learning

15 Brain Area Function VC: visual signal processing (receptive fields) IT: object recognition SP: location recognition HIPP: associative memory PFC: motivation, attention, context

16 Behavioral Data (once again)

17 Model Outline We intended to build a model in order to compare the 1, activities with the fmri data; 2, the preformance with the behavioral data.

18 The retina: sample images The retinal images we use are 8x8 pixel sized, random images placed on a 16x16 pixel arena. The 9 different positions are overlapping.

19 The visual cortex Recieves input from retina Use receptive fields to process images. (2 4 1 possible patterns can be detected at each location of the retina.) Sends processed data to the IT and SP

20 The visual system

21 Inferior Temporal Cortex (IT) Processed Image Location Neglected Encoding Created Input Output IT: stores representation of the different objects as discrete, stable attractors in an RNN.

22 Superior Parietal Cortex (SP) Processed Image Object Neglected Encoding Created Input Output

23 The hippocampal model General hippocampal episodic memory model based on Rolls,E.T. (1996) A theory of hippocampal function in memory. Hippocampus 6:

24 Learning Before the experiment, we initialize our network by storing different number of objects in the recurrent network of IT; random synaptic matrices, modelling associations not relevant for the current context.

25 DG - competitive network a i dg = j w ji sp2dg rj sp + r i dg = F (a dg, sp dg ) k w ki it2dgr k it w ij sp2dg = αrj dg (ri sp w ij sp2dg ) w ij it2dg = αrj dg (ri it w ij it2dg ) Competitive network in the Dentate Gyrus, to make unique, orthogonal representation for each object-location pair. Cortical signals (from SP and IT) arrive to the hippocampus through the entorhinal cortex.

26 CA3 - heteroassociative stage a i ca = j w ji dg2ca rj dg r i ca = F (a ca, sp ca ) w ij sp2ca = αrj ca(r i sp w ij sp2ca ) The strong mossy fiber synapses act as teacher signal for CA3 pyramidal neurons. Perforant path synapses are modified via Hebbian learning.

27 Back to the cortex - heteroassociative stage w ij ca2it = αrj it (ri ca w ij ca2it ) The hippocampal representations are associated to the representation of the original object in the IT.

28 Recall During the recall connections are not modified. The attractor network in the IT help the recall by converging to one of the learned objects.

29 Results

30 Results Control subjects Schizophrenic patients Performance Parameters: r.dg= r.ca3, r.it= n.patterns=550 Performance Parameters: r.dg= r.ca3, r.it= 0.02 n.patterns= Trial number Trial number

31 Results Control subjects Schizophrenic patients Performance Parameters: r.dg= r.ca3, r.it= n.patterns=550 Performance Parameters: r.dg= r.ca3, r.it= 0.02 n.patterns= Trial number Trial number

32 Much left to be done fmri analysis: estimatation of functional connectivities for normal and patholoical situations the inclusion of prefrontal cortex into the model the role of the reduced NMDA-related plasticity. Kinetic model of the drug-altered glutamate-nmda receptor interaction: test and design of drugs