Role of computational studies in neuropharmacology drug designing

1. COMPUTATIONAL
Neuropharmacology-
dynamical
approaches in drug
discovery

2. Word cloud from the 2015
Computational Neuroscience Symposium

3. The biological neuron
A typical neuron consists of a cell
body, dendrites, and an axon.
A neuron is an electrically excitable
cell that processes and transmits
information through electrical and
chemical signals.

4. Neural circuits
Neurons never function in isolation;
they are organized into ensembles
or circuits that process specific
kinds of information.
Afferent neurons, efferent neurons
and interneurons are the basic
constituents of all neural circuits.

5. Brain, body & behavior
The human brain is made of
≈86 billions neurons
Each neuron is connected to
≈10,000 other neurons (average)
1mm3 of cortex contains ≈1 billion
connections

6. To model the brain
T
o emulate → new algorithms
(e.g. deep learning)
To heal → new therapies
(e.g. deep brain stimulation)
T
o understand→ new knowledge
(e.g. visual attention)

7. What kind of models?
Connectionist models for
performances & learning
Biophysical models for simulation
& prediction
Cognitive models for the
emulation of behavior

8. How to build models?
Basic material
•Anatomy and physiology
•Experiments & recordings
•Pathologies & lesions
Working hypotheses
•Extreme simplifications
•Parallel & distributed computing
•Dynamic systems & learning
Validation
•Predictions
•Explanations

9. Computational approaches
 Computational approaches that adopt dynamical models are
widely accepted in basic and clinical neuroscience research as
indispensable tools with which to understand normal and
pathological neuronal mechanisms.
 Although computer-aided techniques have been used in
pharmaceutical research (e.g. in structure- and ligand-based drug
design), the power of dynamical models has not yet been
exploited in drug discovery.
 We suggest that dynamical system theory and computational
neuroscience – integrated with well- established, conventional
molecular and electrophysio- logical methods – offer a broad
perspective in drug discovery and in the search for novel targets
and strategies for the treatment of neurological and psy- chiatric
diseases.

10. Integrative neuropharmacology
 Dynamical system theory offers a unique perspective with which to
investigate the changing activity of single neurons, neural networks and
neural centers.
 'Computational neuropharmacology' combines state-of-the-art
computational techniques that were established in basic neuroscience with
traditional neuropharmological methods.
 Neuroscience could lead to the development of new therapies for
neurological and psychiatric disorders.
 The concept of neurological and psychiatric diseases as dynamical
diseases is then introduced. Integrating potential drug effects into
computational models of neural networks is discussed.
 We highlight two potential applications of computational neuroscience in
drug discovery. These include finding new targets and methods for seizure
control.

11. Computational models: from basic
neuroscience to neuropharmacology
 Computational neuroscience is a relatively new discipline that adopts
dynamical models in basic and clinical science.
 The main goal of computational neuroscience is to build biologically
realistic mathematical models and to test how they match and predict
experimental data by using computer simulations.
 The successes of computational neuroscience were boosted by single-
cell recordings and were theoretically strengthened by elaborated
versions of the Hodgkin–Huxley equations.
 Computer-modeling simulation softwares have been developed to
analyze and interpret experimental findings.
 The availability of electroencephalography (EEG) and
magnetoencephalography (MEG) data that reflect spatio- temporal
neural activity, together with hemodynamic imaging such as positron
emission tomography (PET) and functional magnetic resonance
imaging (fMRI) obtained from human subjects, fostered the emergence
of new modeling techniques that integrate different phenom- ena
occurring at different spatiotemporal scales

12. Various available
models

13. Essentially, all models are wrong
but some are useful.
(George E.P)
Artificial neuron are over-simplified
models of the biological reality, but
some of them can give a fair
account of actual behavior.

14. The biological synapse
Excitatory synapses excite
(depolarize) the postsynaptic cell
via excitatory post-synaptic
potential (EPSP)
Inhibitory synapses inhibit
(hyperpolarize) the postsynaptic
cell via inhibitory post-synaptic
potential (IPSP)

15. Neural circuits

16. Temporal coding
• Rank order (Thorpe & Gautrais, 1991)
→ most of the information about a new
stimulus is conveyed during the first 20 or
50 milliseconds after the onset of the
neuronal response

17. Between cells and tissue
The number of neurons and
synapses in even a small piece of
cortex is immense. Because of this
a popular modelling approach
(Wilson & Cowan, 1973) has been
to take a continuum limit and study
neural networks in which space is
continuous and macroscopic state
variables are mean firing rates.

18. Visual attention
(Vitay & Rougier, 2008)
Several studies suggest that the
population of active neurons in
the superior colliculus encodes
the location of a visual target to
foveate, pursue or attend to.

19. A clockwork orange
Using the output of the focus map
we can control a robot.

20. Plasticity
Synaptic plasticity is the ability of
synapses to strengthen or weaken
over time, in response to increases
or decreases in their activity
Structural plasticity is the
reorganisation of synaptic
connections through sprouting or
pruning.
Intrinsic plasticity is the persistent
modification of a neuron’s intrinsic
electrical properties by neuronal or
synaptic activity

21. Learning
• The cerebellum is specialized for
supervised learning, which is guided
by the error signal encoded in the
climbing fiber input from the inferior
olive learning
• The basal ganglia are specialized for
reinforcement learning, which is
guided by the reward signal encoded
in the dopaminergic input from the
substantia nigra
• The cerebral cortex is specialized for
unsupervised learning, which is
guided by the statistical properties of
the input signal itself, but may also
be regulated by the ascending
neuromodulatory inputs

22. A model of area 3b
(Detorakis & Rougier
, 2013)
Using a neural field, we’ve
modelled the primary sensory
cortex (3b) in the primate using
unsupervised learning.
S timulus
Surface
200 microns
shift
Random dot pattern
Receptive Field
Surface
C
2.0 mm
B
5 mm
40 mm
A
Stimulus
Input Layer
Receptor
Cortical Layer
wi(x)
we(x)
Neuron
wf

23. Decision making
(Topalidou et al, 2016)
Striatum
Cognitive 4 units
GPe
Cognitive 4 units
GPi
Cognitive 4 units
Thalamus
Cognitive 4 units
STN
Cognitive 4 units
Cognitive loop
Cortex
Cognitive 4 units
Striatum
Motor 4 units
GPi
Motor 4 units
GPe
Motor 4 units
Thalamus
Motor 4 units
STN
Motor 4 units
Motor loop
Cortex
Motor 4 units
Associative loop
Task
Environment
Cue
Positions
Cue
Identities
Substantia
nigra pars
compacta
Substantia
nigra pars
compacta
COMPETITION
dopamine
dopamine
reward reward
Lesion
sites
RL
HL
Cortex
Associative
4x4 units
Striatum
Associative
4x4 units
EXT EXT EXT EXT
COMPETITION
CN
+
+
+
+
+
–
–
P
Brain stem structures
(e.g., superior colliculus, PPN)
pr
pc
–
–
Prefrontal
cortex
Premotor cortex
Motor
cortex
Parieto-
temporo-
occipital
cortex
Thalamus
Globus
pallidus
Striatum
Cerebellum
Amygdala
Subthalamic
nucleus
VTA
+
Ventral
pallidum
Saline or muscimol injection
into the internal part of
the Globus Pallidus (GPi)
15 minutes before session
Rewa r d (juice) delivered
according to the reward
probability associated
with the chosen stimulus
P = 0.75
P = 0.25

24. Clock-driven vs event-driven simulation
There are two families of algorithms
for the simulation of neural networks:
• synchronous or clock-driven
algorithms, in which all neurons are
updated simultaneously at every
tick of a clock
• asynchronous or event-driven
algorithms, in which neurons are
updated only when they receive or
emit a spike

25. NEURON
www.neuron.yale.edu/neuron
NEURON is a simulation
environment for modelling
individual neurons and networks of
neurons. It provides tools for
conveniently building, managing,
and using models in a way that is
numerically sound and
computationally efficient.

26. NEST simulator
www.nest-simulator.org
NEST is a simulator for spiking
neural network models that
focuses on the dynamics, size and
structure of neural systems rather
than on the exact morphology of
individual neurons.

27. Brian simulator
briansimulator.org
Brian is a simulator for spiking
neural networks available on
almost all platforms. The
motivation for this project is that a
simulator should not only save the
time of processors, but also the
time of scientists.

28. Reproducible Science
Over the years, the Python language
has become the preferred language
for computational neuroscience.
PyNN is an interface that make
possible to write a simulation script
once, using the Python programming
language, and run it without
modification on any supported
simulator.
rescience.github.io

29. Neurological and psychiatric disorders
as dynamical diseases
 The term ‘dynamical diseases’ was coined because many neurological
and psychiatric disorders are characterized by sudden changes in
qualitative dynamic behavior.
 Techniques applied to analyzing EEG and MEG data also helped to
spread the perspective of dynamical systems. For example, the onset of
seizures was explained by computational models used to interpret
transitions between normal and epileptic activity states of the brain.
 There is some indication, although the clinical relevance is still debated,
that dynamical models can predict seizure development and, thus, the
administration of antiepileptic drugs could be designed accordingly.
 Parkinson’s disease is also a dynamical disease and a method has been
developed to detect preclinical tremor in patients.
 Calculations based on dynamical theories indicate that patients with
Parkinson’s disease display EEG series with a greater complexity than
do normal subjects during the performance of complicated motor tasks.

30.  There are many other diseases in which the concept of dynamical
disease can be applied. Correspondence between clinical and
electrophysiological temporal dimensions has been found in patients
with depression.
 Analysis of scalp EEG data and functional computational models have
been used to study the level of functional connectivity among brain
regions and the synchronization of neural activity in schizophrenia and
Alzheimer’s disease respectively.
 The spatiotemporal development of processes in migraine (i.e. the
appearance of migraine aura and headache associated with visual and
other distortions) has been captured and studied using reaction–
diffusion dynamical models.
 To understand these diseases better, the models of brain disorders
should be refined and the effects of potential drugs (e.g.
neuromodulators that can exert diverse effects on the dynamic
behavior of neuronal elements) should be included.

31. Models of neuromodulatory effects:
molecular screening versus drug design
 Neuromodulators can alter either the intrinsic properties of
neurons or the efficacy of synaptic transmission. Therefore,
rhythmic single-cell and network electrical activity patterns
can also be subject to modulatory effects.
 The pharmacological modification of fast oscillations in the b
and g bands (15–80 Hz) has been reviewed . The
disruption of fast oscillations by neuromodulators has a role
in anesthesia and in cognitive or motor disorders.
 Among asset of psychotropic drugs(chemical substances
that alter brain function, resulting in temporary changes in
behavior, consciousness or mood), noradrenaline-reuptake
inhibitors enhanced q (w7 Hz) and g (w40 Hz) oscillations
in the septo hippocampal system, whereas a selective
serotonin-reuptake inhibitor did not have such an effect.

32.  Allosteric modulators (i.e. compounds that bind to the
regulatory site of a protein to control its binding site)
can influence (i.e. enhance or suppress) electrical
rhythmicity by altering synaptic transmission.
However, the mechanisms of these effects are
unclear.
 The development of computational methods that
integrate conductance-based techniques for
compartmental modeling and a detailed kinetic
description of the pharmacological modulation of
transmitter receptor interactions to test the electro-
physiological and behavioral effects of potential drugs
might be a conceptually new perspective for
computational molecular screening and design

33. Screening and designing procedure

35. Towards a computational
neuropharmacology
 Dynamical system theory, which is a valuable tool in
basic and clinical neurosciences, is under-represented in
main- stream pharmacological research.
 The emergence of a research field, which we propose be
named computational neuropharmacology, would help
to integrate the molecular level and the system-level
descriptions of the nervous system and provide detailed
analysis of the effects of putative drugs on global
electrical patterns and behavioral states.
 Systems pharmacology, as a new perspective towards
this direction, has been reviewed recently .

36. Conclusion
 The proposed methodology is feasible for:
 (i) studying modulatory processes that determine normal
and pathological neural rhythms;
 (ii) providing new strategies for the molecular screening
of putative drugs; and
 (iii) searching for novel target-specific drugs that can
‘tune’ the spatiotemporal neural activity patterns, thus
shifting the system from pathological dynamical states to
normal ones.
 The suggested computational approach is cost effective
compared with traditional methods and it can be
applied to several neurological and psychiatric
diseases.

37. Noise and decision
dx/dt =a(1−x)+(x−y)(1−x), x>0
dy/dy =a(1−y)+(y−x)(1−y), y>0

38. References
 Aradi I, Érdi P. Computational neuropharmacology. Trends in
pharmacological sciences. 2006 May 1;27(5):240-3.
 Ekins S, Mestres J, Testa B. In silico pharmacology for drug discovery:
applications to targets and beyond. British journal of pharmacology. 2007
Sep;152(1):21-37.
 Nair PC, Miners JO. Molecular dynamics simulations: from structure
function relationships to drug discovery. In silico pharmacology. 2014
Dec;2(1):1-4.
 Harrold JM, Ramanathan M, Mager DE. Network‐based approaches in drug
discovery and early development. Clinical Pharmacology & Therapeutics.
2013 Dec;94(6):651-8.