Content uploaded by Jan Treur
Author content
All content in this area was uploaded by Jan Treur on Mar 13, 2024
Content may be subject to copyright.
Cognitive Systems Research 83 (2024) 101177
Available online 11 October 2023
1389-0417/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Higher-order adaptive dynamical system modeling of the role of epigenetics
in anxiety disorders
☆
Shivant Kathusing, Natalie Samhan, Jan Treur
*
Vrije Universiteit Amsterdam, Department of Computer Science, Social AI Group, Netherlands
ARTICLE INFO
Action editor: Alexei V. Samsonovich
Keywords:
Anxiety disorder
Epigenetics
Temporal-causal network
ABSTRACT
In this paper, a fth-order adaptive self-modelling network model is introduced to describe epigenetic
involvement in the development of anxiety disorders and its regulation by a possible epigenetics-based thera-
peutic method. Multiple orders of adaptivity are used in the model to depict the development process, where a
higher pathway of any order of adaptivity adapts characteristics of pathways in lower orders and acts as a form of
control. These orders of adaptivity and their interlevel interaction were modelled as a higher-order adaptive
dynamical system according to the self-modelling network modelling principle. The model was inspired by the
structure of the relevant human biological and neurological processes. In addition to modelling the development
of an anxiety disorder, also the possibility of an epigenetics-based therapy is suggested and computationally
analyzed in this paper.
1. Introduction
According to the World Health Organization, 3.6 percent globally
have an anxiety disorder. Although anxiety disorder is widely known,
not everyone receives an appropriate treatment for it. Such treatments
involve medication, counselling or Cognitive Behavioural Therapy
(CBT). Recently, researchers discovered that there exists an anxiety gene
which can downregulated by epigenetics (Mucha et al., 2023). Anxiety
disorder, which includes generalized anxiety disorder, social anxiety
disorder (social phobia), specic phobias and separation anxiety disor-
der, is a mental health condition characterized by persistent, excessive
and uncontrollable fear or worry that may obstruct a person’s life. This is
not the same phenomenon as the normal innate experience of anxiety
that most people encounter. For example, being worried or anxious
before an important presentation or an exam, is natural, whereas anxiety
disorders imply an irrational and overwhelming sense of fear that persist
when there is no sign of immediate danger or threat. Furthermore, a
patient’s thoughts and emotions are disrupted due to the anxiety expe-
rienced in anxiety disorders. This often leads to impairment and distress.
Chronic stress and past or childhood experiences are one of the factors
that may induce the development of anxiety disorder (An et al., 2022;
Khan & Khan, 2017; Lin & Tsai, 2020; Meaney, 2010; Rosa, Formolo, Yu,
Lee, & Yau, 2022; McEwen, 2017). One of the reasons people develop
anxiety disorder is chronic stress. For the models described in the paper,
it is assumed that in a patient with anxiety disorder chronic stress dis-
rupts the function of the pre-frontal cortex for regulation of stress (Gold
et al., 2016). In this way, psychological trauma can cause changes in the
amygdala activation within the brain. These changes can lead to high
levels of anxiety and stress issues (Bartlett et al., 2017).
In this paper, the research question ‘How can the role of epigenetics
in anxiety disorders be modelled?’ is being addressed. A fth-order
adaptive dynamical system model has been suggested in order to
model to role of epigenetics in anxiety disorder where multiple levels of
adaptation are considered when analyzing the anxiety regulation.
Moreover, this paper also explores a possible therapy method based
on the studies by Fu et al. (2021), Mucha et al. (2023) and Scott et al.
(2015). These studies focus on patients diagnosed with a disorder or
disease whose miRNA has been manipulated, specically through
intravenous injections in cancer patients. Signicant progress has been
achieved in the eld of cancer research, and therapies utilizing miRNA
(microRNA) are already being studied in clinical settings. However,
when it comes to treating neurological disorders, like anxiety disorder,
the utilization of miRNA on humans is still in its early stages of devel-
opment. The models for the adaptive dynamical systems addressed in
this paper are designed as self-modelling adaptive network models,
based on the Network-Oriented Modelling approach described in (Treur,
☆
This paper was proposed for the CSR Special Issue VSI:FICS by the BICA*AI’23 PC.
* Corresponding author.
E-mail addresses: kathusingshivantzakelijk@gmail.com (S. Kathusing), natalie.samhan@ing.com (N. Samhan), j.treur@vu.nl (J. Treur).
Contents lists available at ScienceDirect
Cognitive Systems Research
journal homepage: www.elsevier.com/locate/cogsys
https://doi.org/10.1016/j.cogsys.2023.101177
Received 14 September 2023; Accepted 30 September 2023
Cognitive Systems Research 83 (2024) 101177
2
2016, 2020). In Section 2, the background of anxiety disorder is elab-
orated. In Section 3, the model for the development of anxiety disorder is
presented. In Section 4 simulations are shown for the concerning model.
2. Background anxiety disorder and epigenetics
Fear and anxiety are innate emotions that occur when the brain
senses danger or (potential) conicts. These emotions are used since the
existence of humans as an essential mechanism for survival. However,
anxiety disorders do not concern this temporary fear or worry. An
epigenetically based disorder is an abnormality in the functioning
regulation via expression of one or more genes. This can either arise
congenital or by an acquired state, where someone passes for example, a
stressful childhood. Epigenetic disorders usually concern epigenetic
factors relating to external stimuli. The condition is usually reversible.
Mucha et al. (2023) have concluded that anxiety disorder is mainly
epigenetic and that it may be epigenetically reversible. Once the
restoring of the expression of genes has successfully been done, the
symptoms disappear. This operation can be achieved through for
example adapted nutrition, intravenous injections, and adapted habits.
Moreover, miRNAs, short for microRNAs, are small RNA molecules
(usually around 22 nucleotides long) that function in regulating gene
expression. They achieve this by binding to messenger RNA (mRNA)
molecules, leading to their degradation and, consequently, repression of
translation (Pe˜
na & Nestler, 2018; Penner-Goeke & Binder, 2019; Rosa
et al., 2022). MicroRNAs are an interface between genes and the envi-
ronment (Scott et al., 2015). The signicance of miRNAs goes beyond
their role as developmental switches and regulators of cellular differ-
entiation. They also have a more intricate function in ne-tuning gene
expression by reducing the abundance of specic mRNA transcripts,
without completely eliminating them. This paper focuses of the miRNA
called miR-483-5p, as researchers found that miR-483-5p in the amyg-
dala of mice represses the stress-gene PGAP2, which is responsible for
anxiety-like behavior (Mucha et al., 2023). When the mice in the
experiment experienced stress, miR-483-5p is upregulated in the syn-
apses of the neurons in the amygdala. This upregulation causes the
dendritic arbor to become compact and lopodia evolve into dendritic
spines. This experiment shows that the miR-483-5p acts as a guard in
response to the body’s stress and thus has an anxiolytic effect. Due to the
challenges involved in directly studying the role of miRNA in anxiety
disorders in humans, researchers frequently rely on animal models.
While it is not possible to replicate the entire range of symptoms seen in
human anxiety disorders, animal models provide valuable insights into
the mechanisms that contribute to individual vulnerability and the
progression of these conditions. By using animal models, researchers are
able to gain a better understanding of the underlying mechanisms and
factors involved in the development of anxiety disorders.
3. Modeling adaptive dynamical systems by networks
The processes described above involve interaction of multiple types
of dynamic and adaptive processes. Together they form a higher-order
adaptive dynamical system. In this paper, to model this the self-
modeling network modelling approach introduced by (Treur, 2020)
has been used, as it has been shown that this is a suitable approach to
model higher-order adaptive dynamical systems: any smooth higher-
order adaptive dynamical system has a canonical representation as a
self-modeling network (Hendrikse et al., 2023; Treur, 2021).
Temporal-causal network models as introduced in (Treur, 2016) are
dened by three types of network characteristics: connections
(describing causal impacts) from node X to node Y with their weights
ω
X,
Y
(connectivity characteristics), combination functions c
Y
to aggregate the
causal impacts on one node Y (aggregation characteristics), specied by
combination function weights γ
i,Y
and combination function parameters
π
i,j,Y
, and speed factors
η
Y
indicating the speed by which a state Y
changes upon aggregated impact (timing characteristics). The nodes X in
temporal-causal networks are also called states as they are dynamic with
activation values X(t) that change over time. Their connections are
interpreted as causal relations that affect the activation levels of other
states. For the network’s dynamics, the above network characteristics
are incorporated into a canonical difference equation used:
Y(t+Δt) = Y(t) +
η
Yc
π
Y,Y
ω
X1,YX1(t),⋯,
ω
Xk,YXk(t)) − Y(t)]Δt(1)
where X
1
to X
k
represent states from which Y receives incoming con-
nections. Equation (1) bears a resemblance to the format of recurrent
neural networks.
The software environment outlined in Treur (2020a, Ch. 9) includes
a library of approximately 70 basic combination functions for use in the
model design process. Examples of these functions used here are listed in
Table 1. Note that when the function alogistic
σ
,
τ
(V
1
, …,V
k
) is applied to
intended positive values, it is cut off at 0 if its formula produces a
negative value. However, this cut-off is not applied when the function is
applied to intended negative values. Overall, these concepts enable the
declarative design of network models and their dynamics based on
mathematically dened functions and relations. By instantiating this
general difference equation (1) by proper values for the network char-
acteristics for all states Y, the software environment runs a system of n
difference equations where n is the number of states in the network.
By the self-modeling principle (also called reication principle)
introduced in (Treur, 2020a), any of the above network characteristics
can be made adaptive by adding a self-model state for it to the network.
For the model introduced here, self-model states W
X,Y
for connectivity
have been used that represent connection weights
ω
X,Y
, so that the
network can model itself concerning its own connectivity. These self-
models are a part of the temporal-causal network, they are placed at a
level higher than the base network. These self-model states for con-
nectivity have incoming connections for causal impacts on it and out-
going connections to effectuate them. Effectuating takes place by using
the value W
X,Y
(t) for
ω
X,Y
in (1). For example, when all connection
weights
ω
Xi,Y are made adaptive by adding self-model states WXi,Y for
them, equation (1) becomes
Y(t+Δt) = Y(t) +
η
Yc
π
Y,YWX1,Y(t)X1(t),⋯,WXk,Y(t)Xk(t)) − Y(t)]Δt
(2)
As the self-modeling principle can also be applied to self-model states,
higher-order self-model states can occur. For example, if a context state c
has impact on a self-model state W
X,Y
, and the weight
ω
c,Wx,y
of the
causal connection from c to W
X,Y
is adaptive as well, then this can be
modeled by a second-order self-model state W
c,Wx,y
. This still can be
iterated for higher orders. In Section 4, this has been applied obtaining
up to fth-order self-model states which distinguish different orders of
adaptation that play a role in the processes.
4. The higher-order adaptive self-modelling network model
To conceptualize the higher-order adaptive dynamical system based
on the biological and mental processes of the development of anxiety
disorder due to stress, which eventually becomes chronic stress, a self-
modelling adaptive network model, as described in (Treur, 2020), was
used. The different orders of plasticity that describe the dynamics of the
brain’s development of anxiety disorder are using up to ve adaptation
levels. The base level represents an emotion regulation process as
described in (Treur, 2016), Section 3.3, and provides a mechanism that
tries to suppress the anxiety. In this mechanism, a control state repre-
senting a part of the prefrontal cortex detects whether an undesired level
of emotion occurs and if so tries to suppress this. This will fail when the
prefrontal cortex’s activation is too weak. The adaptations are driven by
changing environmental circumstances which lead to changes in certain
pathways.
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
3
4.1. The overall structure of the network model
From a broad perspective, evolutionary processes can often be seen
as adaptation processes that modify the physical world by establishing
new causal pathways modulating or inhibiting existing ones. These
changes involve altering the strength of causal connections in these
existing pathways, ranging from very low to high or vice versa. Such
adaptations are driven by shifts in environmental conditions, favouring
organisms with more advantageous causal pathways, which subse-
quently become dominant in the population (Fessler et al., 2015). These
environmental conditions play a crucial role in shaping these adapta-
tions. This concept of causal pathways modulating other (existing)
causal pathways aligns with a self-modeling network architecture,
where connection weights in causal pathways are represented by self-
model states according to subsequent levels (Treur, 2020), Ch 7, Sec-
tion 7.3.1. Following this perspective, the following ve orders of
adaptation have been distinguished:
•First-order adaptation
The maintenance and strengthening of connections between neurons
in the brain, which is regulated by enzymes.
•Second-order adaptation
The enzyme that is responsible for the connections involved in the
anxiety emotion is being produced, with causal pathways positively
affecting the causal pathways for feeling anxiety. This is regulated by
mRNA.
•Third-order adaptation
mRNA is produced that is responsible for carrying genetic informa-
tion to the cytoplasm from the DNA for protein synthesis to take
place, with causal pathways positively effecting the causal pathways
for the enzyme. This is regulated by certain genes in the DNA.
•Fourth-order adaptation
PGAP2 is the gene that contains the genetic information, with causal
pathways controlling the causal pathways for mRNA production.
•Fifth-order adaptation
miR-483-5p is responsible for regulating the expression of the PGAP2
gene, with causal pathways negatively affecting the causal pathways
for PGAP2.
For a picture of these levels, see Fig. 1. The upper level (fth self-
model level), which contains the ‘epigene’ miR-483-483-5p, causes the
lower levels to not operate properly by the effect of adaptation when this
Table 1
The combination functions used in the introduced network model.
Notation Formula Parameters
Advanced logistic sum alogistic
σ
,
τ
(V
1
, …,V
k
) 1
1+e−
σ
(V1+⋯+Vk−
τ
)−1
1+e
στ
(1+e−
στ
Steepness
σ
Excitability threshold
τ
Steponce steponce
α
,β
(V) 1 if
α
≤t ≤β
0 else
Time t Start
α
End β
Fig. 1. Graphical representation of the network model for development of an anxiety disorder.
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
4
level’s working is being disturbed.
4.2. The detailed conceptual representation of the network model
In order to visualize the processes and interactions in detail of the
case in question, a graphical representation of the temporal-causal
network has been made rst. The graphical model in Fig. 1 contains
the rst part of the following characteristics:
•the states X,Y
•the connections between states X → Y
•the connection weight (connectivity)
ω
X,Y
Table 2
All states and their explanation. The colours indicate the levels for different adaptation orders from base level to fth-order self-model level.
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
5
•the aggregation of multiple impacts (combination function) on a
state c
Y
•the timing of the effect (speed factor) of causal impact
η
Y
For an overview of all states and their explanation, see Table 2.
The states can take on activation values between 0 and 1 that may
vary over time. The role W, for self-modelling of connection weights, is
used for the self-model states. The ovals in the model in Fig. 1 represent
nodes and the connections between notes is indicated as arrows with the
end being the direction of the causal impact.
There are different types of states:
•context states con: an external stimulus or supporting environment,
e.g. con
stress
for representing the stimulus stress
•sensor states ss: sensing of the stimulus, e.g. ss
s
for sensing the
stimulus s for stress
•sensory representation states srs: mental representation, e.g. srs
s
mental representation for the stimulus s
•preparation states ps: expressing preparation for a body state, e.g. ps
b
for the body state b
•feeling states fs: feeling of the stimulus, e.g. fs
b
for feeling the
emotion/feeling b
•prefrontal cortex states (control states) pfc: regulating the emotion/
feeling, e.g. pfc
b
for regulating the emotion/feeling b
•Self-model weight states W: self-model representation state, e.g.
W
fs
b
,
pfc
b
for the weight of the base level connection from fs
b
for
context to pfc
b
for pre-frontal cortex/regulating state
The outgoing downward causal connections (pink downward ar-
rows) in Fig. 1 represent the specic causal impact each of these self-
modelling state has. The upward (or leveled) causal connections (blue
arrows) to the self-modelling states give them the dynamics as desired.
They are used to specify, together with the combination function that is
chosen and the downward connection, the particular adaptation prin-
ciple that is addressed. The dotted arrows represent a negative
connection X → Y, with X supressing Y. Positive connections are rep-
resented by solid black arrrows.
4.3. The detailed representation of the network model for the development
of anxiety disorder
In order to obtain a fully specied detailed design of the network
model that also can be executed in MATLAB, based on the above con-
ceptual representation, a detailed representation in terms of role
matrices as described in (Treur, 2020) has been made. See Appendix A
for the full specication of these role matrices. The following role
matrices are included:
Connectivity characteristics
▪ mb: base connectivity
▪ mcw: connection weights
ω
Aggregation characteristics
▪ mcfw: combination function weights γ
▪ mcfp: combination function parameters
π
Timing characteristics
▪ ms: speed factors
η
▪ iv: initial values of the states
Each row represents a state X
i
. The pink cells in each row of the role
matrices indicate which other states affect X
i
from that role. The green
cells represent values affecting X
i
. For example, in mb the pink cells
represent the states, the state X
i
of the row has an incoming connection
from. In role matrix mcw, the green cells indicate the weight of the
connection between the row state and the connected state in that same
cell which is indicated in mb. The pink cells in mcw indicate which
states X
j
play the role of self-model state for the weight of the connection
indicated in that same cell in mb.
4.4. The conceptual and detailed representation of the network model for
a therapy for anxiety disorder
A second case that has been modelled, describes the effect of a
therapy as suggested by Mucha et al. (2023). For a conceptual picture,
see Fig. 2. This natural therapy consists of suppressing the gene PGAP2
with the help of the epigenetically activated non-coding RNA miR-483-
5p, by intravenous injecting miR-483-5p. This will have an anxiolytic
effect. For this, in the conceptual representation a world state con
therapy
has been added to Table 2:
See also in the green upper plane in Fig. 2. This world state acts as an
intravenous injection of miR-483-5p. For all role matrices, see the Ap-
pendix A section.
5. Simulation experiments
In order to illustrate the computational models of the role of epige-
netics in anxiety disorders and a therapy for anxiety disorders, two
scenarios have been created. In the rst scenario, a person develops
anxiety disorder as the person is exposed to chronic stress. In the second
scenario an assumed epigenetic therapy by the use of intravenous in-
jection of miRNA miR-483-5p is presented. In order to obtain the
simulation processes of these scenarios, the dedicated software
described in (Treur, 2020), Ch 9 was used. In both scenarios there is a
baseline level of anxiety introduced that remains steady at level 0.5. To
model the development of anxiety disorder symptoms, a chronic stress
factor in addition to that has been introduced that is consistent in both
scenarios (by the step-once function).
5.1. Simulation for Scenario 1: The development of anxiety disorder
In the beginning of the simulation, the W-states go up, because of the
activation state con
stress
; see Fig. 3 This activation state is needed to
activate the brain’s regulation processes. Without stress there is no need
for the miR-483-5p to go up. That is why the con
stress
has a constant level
of 0.5 from the beginning to the end of the case. W
fs
b
,pfc
b
and W
ps
b
,pfc
b
are
overlapping each other, because they have the same values. The sensor
state ss
s
and sensory representation state srs
s
gradually come up as re-
action to the world state (stimulus) con
stress
and afterwards becomes
steady between time points 50 and 120. This can be interpreted as a
person constantly feeling stress for a certain time. The preparation state
for anxiety slowly moves up, but then becomes steady as srs
s
, fs
b
and pfc
b
are steady as well. The person in question has no anxiety disorder yet, as
anxiety is an innate emotion to stress (See introduction). When chronic
stress con
chronicstress
is sensed from time point 120 to 280, the sensor
state ss
s
and sensory representation state srs
s
quickly move up. The
person in question senses a lot of chronic stress. The ps
b
and fs
b
now
move up as reaction to the ss
s
and srs
s
moving up. These mechanisms
move the W-states down, with W
con
DNA
, W
con
mRNA
, W
con
enzyme
, W
con
connection
, W
being the rst one, because of the negative upward connections from ps
b
and fs
b
. The ps
b
and fs
b
rstly move the pfc
b
a little up. The activity of the
prefrontal cortex of the person in case now increases, because of the
regulative mechanisms as response to the amygdala state fs
b
for feeling
anxiety. The activity of the prefrontal cortex state pfc
b
goes eventually to
zero, because the weight adaptation W-states for the connections to pfc
b
become weaker. The latter mechanism comes from the fth-order self-
model level through the downward pink arrows via the lower-order self-
model levels. At the fth-order level, the miRNA miR-483-5p becomes
low, which means that it cannot suppress the anxiety gene PGAP2
anymore. Subsequently, there is a dysregulation in the mRNA and the
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
6
enzymes that are involved in anxiety behavior. This results in the W-
states at all levels moving down. As a result, the patient has no control
anymore of the amygdala’s response via the prefrontal cortex and has
now developed anxiety disorder. From 280 time units it can be observed
that the prefrontal cortex’s level is still zero and the patient is left with
constantly feeling anxiety, even when there is no excessive chronic stress
anymore. The brain’s functioning of this person has been damaged after
the episode between 120 and 280 time units.
5.2. Simulation for Scenario 2: Epigenetic therapy for anxiety disorders
The second case that has been modelled, describes the effect of
natural therapy as described in the paper of Mucha et al. (2023); see
Fig. 4. This natural therapy consists of suppressing the gene PGAP2 with
the help of the epigenetically activated non-coding RNA miR-483-5p, by
intravenous injecting miR-483-5p. This will have an anxiolytic effect. In
the conceptual representation of this case an extra world state con
therapy
has been added. This world state acts as an intravenous injection of miR-
483-5p. See Section 4.1 for the explanation about the development of
anxiety disorder, which is also shown in Fig. 3.
For the therapy, con
therapy
is being upregulated from 450 to 610 to
level 1.0. This signies miR-483-5p being intravenous injected in the
patient. After the miRNA has been upregulated in the person, the
adaptive levels all go up, which represents the functioning of the
mechanisms in the brain. The person does not feel persisting stress
anymore. This is depicted by fs
b
and ps
b
going down to normal levels
between 0 and 0.1 as before the development of anxiety disorder. Also
the ss
s
and the srs
s
are back to previous constant levels between 0.37 and
0.5.
The weight self-model state W
con
DNA
, W
con
mRNA
, W
con
enzyme
, W
con
connection
,
W has a positive incoming connection from con
therapy
with weight 2,
which means that the manually upregulating has a relatively strong
impact. This connection has been chosen because it is assumed that miR-
483-5p has the most signicant role in regulating the expression of the
anxiety gene PGAP2.
Comparing this simulation to the simulation in Fig. 3 for Scenario 1,
the brain’s regulation function is brought back to normal and stays
functioning. Whereas, in Scenario 1, even after chronic stress is not
present anymore, the brain’s regulation is damaged.
It is evident in the rst scenario that without proper therapeutic
methods to alleviate anxiety disorder symptoms, the anxiety expression
will remain high even after the chronic stress factor is low again.
However, the second scenario demonstrates how after the therapeutic
intervention is introduced, the anxiety expression is regulated and de-
creases again towards the end of the simulation.
6. Discussion
The introduced higher-order adaptive dynamical system model has
had some inspiration by the model for evolutionary development
described in (Treur, 2019), see also (Treur, 2020), Ch 7. The current
model is an integrative computational model as it combines modeling of
biological processes concerning genetics and epigenetics with modeling
of mental processes concerning stress, anxiety and emotion regulation.
The modeling choices made are based on empirically founded literature,
as discussed in Section 2. After nishing the paper we came across the
Fig. 2. Graphical representation of the network model for anxiety disorder and assumed therapy for it.
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
7
interesting new recent paper by Joel Nigg (2023) and used it for a form
of additional validation of our model. Also this new paper conrms the
main choices we made. Examples of this are the following:
•Further evidence for the assumed relation between environment and
gene expression as expressed here:
‘The critical element is that epigenetic remodeling is dynamically
responsive to the environment; it can exemplify experiential pro-
gramming of gene expression (Meany, 2010; O’Donnell & Meaney,
2020).’ (Nigg, 2023), p. 302
•Further evidence to support the choice for dynamical system
modeling in this paper can be found in the paper of Nigg (2023), p.
301:
‘…empirical evidence that psychopathology actually behaves like a
dynamic system has accrued only recently (Kendler et al., 2011;
Kuranova et al., 2020; Lichtwarck-Aschoff et al., 2009; Nelson et al.,
2017; Schiepek et al., 2014, 2016; Wichers et al., 2015). With those
data in support, I stipulate that psychopathology is not a static entity
responding to linear, unidirectional inputs. Rather it emerges from
dynamic system interplay (Miller & Bartholomew, 2020)’ (Nigg,
2023), p. 301
•Further evidence for the choice to consider emotion regulation as a
crucial element of the base level processes can also be found there
(also, see (Nigg, 2017)):
‘I propose self-regulation as a common or shared psychological
mechanism across most of psychopathology. For example, via either
automatic or cognitive routes, anxiety and mood disorders are
related to failure to regulate affect, ADHD to dysregulation of im-
pulse, and schizophrenia to dysregulation of motivation.’ (Nigg,
2023), p. 301
So, also in the most recent literature it is claimed that the analysed
processes form an adaptive dynamical system, so a modeling approach
addressing that is needed. However, besides such claims no actual
earlier computational models are known to the authors that are
Fig. 3. Simulation for Scenario 1.
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
8
integrative in the sense that they integrate all these different types of
processes. The modeling approach chosen here is able to model any
smooth dynamical system as has been shown in (Hendrikse et al., 2023;
Treur, 2021) and has turned out to be a good choice to address this. The
only computational model that has some overlap is the model for
handling sensory processing sensitivity for persons with autism (David
et al., 2023) which was developed in parallel with the model for anxiety
disorders described in the current paper and adopted some of the ele-
ments originally explored for the current paper.
Future research can further investigate ulterior therapy methods
which do not include any form of injection or medication. Specic case
studies taking more context factors into account can shed more insight
and validity into model presented in this paper. In this paper a therapy
based onan intravenous injection was chosen as possible therapy.
However, it is unknown if this will be feasible in terms of availability and
compatibility of the miR-483-5p with patients that suffer from anxiety
disorder. Increasing the weight of the therapy context state to the out-
going connections can be tried out in a model but is is a different thing in
actuality due to the multiple extraneous variables that factor in the
complex nature of one’s physiology. Such therapeutic methods need to
be studied long-term over real life patients to more condently estimate
and map the simulation experiments to real experiments. Also, the
transferability of this medical intervention has mainly been studied for
animals such as mice. Albeit mice having similar brain physiology as
humans, there is still a gap of complexity and interceding factors that
limit the extend in which we can generalize the ndings of the study the
Fig. 4. Simulation for Scenario 2.
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
9
therapeutic method is based on. Furthermore, it would be useful to
explore variations of similar simulation experiments.
Moreover, as usually is the case, for the model presented in this paper
some simplifying assumptions were made concerning development and
treatment of anxiety disorders, mainly following (Mucha et al., 2023).
The regulating effect of miR-483-5p displayed in the model can be
extended to take into account other stress-related genes which are
impacted directly by its regulation such as Gpx3 and Macf1 (Mucha
et al., 2023). An extension of the model can further explore the interplay
between the regulation of miR-483-5p and the methylation patterns of
other key genes related in anxiety such as the NR3C1, FKBP5, and MAOA
genes. Furthermore, studies have shown the hypermethylation of HECA
affects anxiety symptoms differently depending on the sex of the indi-
vidual. One can further take into account context factors such as the sex
of the individual to model the interplay between the associated genes
and the manifested behavior and emotional effect. Ultimately, there are
numerous factors, both biological and environmental that can
contribute to the manifestation of anxiety disorders, and although this
model does not address all of them, it still offers an insightful basis that
studies the factors and their adaptive effects.
Declaration of Competing Interest
The authors declare that they have no known competing nancial
interests or personal relationships that could have appeared to inuence
the work reported in this paper.
Data availability
Data will be made available on request.
Appendix A
The dedicated software environment uses detailed specication of all network characteristics of a network model by role matrices. In each role
matrix (see below), each state has its row where it is listed which are the impacts on it from the role addressed by that role matrix. The base con-
nectivity characteristics are specied by role matrices mb and mcw. Role matrix mb species for each row the other states at the same or lower level
from which the indicated state gets its incoming connections. In role matrix mcw the connection weights are specied. Here nonadaptive connection
weights are indicated in mcw by a number in a green shaded cell. In contrast, adaptive connection weights are indicated in pink-red shaded cells by a
reference to the (self-model) adaptation state representing the adaptive value.
The role matrices ms specify speed factors and initial values are specied in iv. Role matrix mcfw denes the network characteristics for ag-
gregation by indicating the selection of combination functions for all states. Other network characteristics for aggregation are specied by matrix
mcfp for the parameter values of the selected combination functions.
A: Role matrices scenario 1
A: Role Matrices Scenario 1
mb base connectivity 1 2 3
X
1
ss
s
X
6
X
7
X
2
srs
s
X
1
X
3
ps
b
X
2
X
4
X
5
X
4
fs
b
X
3
X
5
X
5
pfc
b
X
3
X
4
X
6
con
stress
X
6
X
7
con
chronicstress
X
7
X
8
W
fsb,pfcb X
10
X
9
Wpsb,pfcb
X
10
X
10
con
connection
X
10
X
11
Wcon
connection
,WX
12
X
12
con
enzyme
X
12
X
13
Wcon
enzyme
,Wcon
connection
,WX
14
X
14
con
miRNA
X
14
X
15
W
con
miRNA
,
W
con
enzyme
,
W
con
connection
,
W
X
16
X
16
con
DNA
X
16
X
17
Wcon
DNA
, Wcon
miRNA
, Wcon
enzyme
,Wcon
connection
,WX
4
X
18
X
3
X
18
con
epigene tics
X
18
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
10
mcw connection weights 1 2 3
X
1
ss
s
1 1
X
2
srs
s
1
X
3
ps
b
1 1 -1
X
4
fs
b
1 -1
X
5
pfc
b
X
8
X
9
X
6
con
stress
1
X
7
con
chronicstress
1
X
8
Wfsb,pfcb X
11
X
9
Wpsb,pfcb X
11
X
10
con
connection
1
X
11
Wcon
connection
,WX
13
X
12
con
enzyme
1
X
13
Wcon
enzyme
,Wcon
connection
,WX
15
X
14
con
miRNA
1
X
15
Wcon
miRNA
, Wcon
enzyme
,Wcon
connection
,WX
17
X
16
con
DNA
1
X
17
Wcon
DNA
, Wcon
miRNA
, Wcon
enzyme
,Wcon
connection
,W-1 1 -1
X
18
con
epigene tics
1
ms speed factors 1
X1sss0.1
X2srss0.1
X
3
ps
b
0.1
X
4
fs
b
0.1
X5pfcb0.1
X6constress 0
X7conchronicstress 1
X8Wfsb,pfcb 0.1
X
9Wpsb,pfcb
0.1
X
10
con
connection
0
X11 Wconconnection,W0.1
X12 conenzyme 0
X13 Wconenzyme,Wconconnection,W0.1
X14 conmiRNA 0
X
15 W
con
miRNA
,
W
con
enzyme
,
W
con
connection
,
W
0.1
X
16
con
DNA
0
X17 WconDNA, WconmiRNA, Wconenzyme ,Wconconnection,W0.05
X
18
con
epigene tics
0
iv initial values 1
X
1
ss
s
0
X
2
srs
s
0
X
3
ps
b
0
X
4
fs
b
0
X
5
pfc
b
0
X
6
con
stress
0.5
X
7
con
chronicstress
0
X
8
W
fsb,pfcb 0
X
9
Wpsb,pfcb
0
X
10
con
connection
1
X
11
Wcon
connection
,W0
X
12
con
enzyme
1
X
13
Wcon
enzyme
,Wcon
connection
,W0
X
14
con
miRNA
1
X
15
W
con
miRNA
,
W
con
enzyme
,
W
con
connection
,
W
0
X
16
con
DNA
1
X
17
W
con
DNA
,
W
con
miRNA
,
W
con
enzyme
,
W
con
connection
,
W
0
X
18
con
epigene tics
1
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
11
mcfw combination function weights algoistic steponce
X
1
ss
s
1
X
2
srs
s
1
X
3
ps
b
1
X
4
fs
b
1
X
5
pfc
b
1
X
6
con
stress
1
X
7
con
chronicstress
1
X
8
Wfsb,pfcb 1
X
9
Wpsb,pfcb 1
X
10
con
connection
1
X
11
Wcon
connection
,W1
X
12
con
enzyme
1
X
13
Wcon
enzyme
,Wcon
connection
,W1
X
14
con
miRNA
1
X
15
Wcon
miRNA
, Wcon
enzyme
,Wcon
connection
,W1
X
16
con
DNA
1
X
17
W
con
DNA
,
W
con
miRNA
,
W
con
enzyme
,
W
con
connection
,
W
1
X
18
con
epigene tics
1
mcfp combination function parameters algoistic Kolom1 Kolom2 Kolom22
function 1 2 1 2
parameter σ τ α β
X
1
ss
s
5 0.5
X
2
srs
s
5 0.5
X
3
ps
b
5 0.8
X
4
fs
b
5 0.6
X
5
pfc
b
5 0.5
X
6
con
stress
5 0.5
X
7
con
chronicstress
120 280
X
8
Wfsb,pfcb 5 0.5
X
9
Wpsb,pfcb 5 0.5
X
10
con
connection
5 0.5
X
11
Wcon
connection
,W5 0.5
X
12
con
enzyme
5 0.5
X
13
Wcon
enzyme
,Wcon
connection
,W20 0.5
X
14
con
miRNA
5 0.5
X
15
Wcon
miRNA
, Wcon
enzyme
,Wcon
connection
,W20 0.5
X
16
con
DNA
5 0.5
X
17
Wcon
DNA
, Wcon
miRNA
, Wcon
enzyme
,Wcon
connection
,W20 0.7
X
18
con
epigeneti cs
5 0.5
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
12
B: Role matrices scenario 2
mb base connectivity 1 2 3 4
X
1
ss
s
X
6
X
7
X
2
srs
s
X
1
X
3
ps
b
X
2
X
4
X
5
X
4
fs
b
X
3
X
5
X
5
pfc
b
X
3
X
4
X
6
con
stress
X
6
X
7
con
chronicstress
X
7
X
8
Wfsb,pfcb X
10
X
9
Wpsb,pfcb
X
10
X
10
con
connection
X
10
X
11
Wcon
connection
,WX
12
X
12
con
enzyme
X
12
X
13
Wcon
enzyme
,Wcon
connection
,WX
14
X
14
con
mRNA
X
14
X
15
Wcon
mRNA
,Wcon
enzyme
,Wcon
connection
,WX
16
X
16
con
DNA
X
16
X
17
Wcon
DNA
,Wcon
mRNA
,Wcon
enzyme
,Wcon
connection
,WX
4
X
18
X
3
X
19
X
18
con
miRNA
X
18
X
19
con
therapy
X
19
mcw connection weights 1 2 3
X1sss1 1
X
2
srs
s
1
X3psb1 1 -1
X4fsb1 -1
X5pfcbX8X9
X
6
con
stress
1
X
7
con
chronicstress
1
X8Wfsb,pfcb X11
X9Wpsb,pfcb X11
X10 conconnection 1
X11 Wconconnection,WX13
X
12
con
enzyme
1
X13 Wconenzyme,Wconconnectio n,WX15
X14 conmRNA 1
X15 WconmRNA, Wconenzyme,Wconconne ction,WX17
X
16
con
DNA
1
X17 WconDNA, WconmRNA, Wconenzyme ,Wconconnection,W-1 1 -1
X
18
con
miRNA
1
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
13
iv initial values 1
X
1
ss
s
0
X
2
srs
s
0
X
3
ps
b
0
X
4
fs
b
0
X
5
pfc
b
0
X
6
con
stress
0.5
X
7
con
chronicstress
0
X
8
Wfsb,pfcb 0
X
9
Wpsb,pfcb 0
X
10
con
connection
1
X
11
Wcon
connection
,W0
X
12
con
enzyme
1
X
13
Wcon
enzyme
,Wcon
connection
,W0
X
14
con
mRNA
1
X
15
Wcon
mRNA
, Wcon
enzyme
,Wcon
connection
,W0
X
16
con
DNA
1
X
17
Wcon
DNA
, Wcon
mRNA
, Wcon
enzyme
,Wcon
connection
,W0
X
18
con
miRNA
1
X
19
con
therapy
0
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
14
ms speed factors 1
X
1
ss
s
0.1
X
2
srs
s
0.1
X
3
ps
b
0.1
X
4
fs
b
0.1
X
5
pfc
b
0.1
X
6
con
stress
0
X
7
con
chronicstress
1
X
8
Wfsb,pfcb 0.1
X
9
Wpsb,pfcb 0.1
X
10
con
connection
0
X
11
Wcon
connection
,W0.1
X
12
con
enzyme
0
X
13
Wcon
enzyme
,Wcon
connection
,W0.1
X
14
con
mRNA
0
X
15
Wcon
mRNA
, Wcon
enzyme
,Wcon
connection
,W0.1
X
16
con
DNA
0
X
17
Wcon
DNA
, Wcon
mRNA
, Wcon
enzyme
,Wcon
connection
,W0.05
X
18
con
miRNA
0
X
19
con
therapy
1
mcfw combi. funct. weights algoistic steponce
X
1
ss
s
1
X
2
srs
s
1
X
3
ps
b
1
X
4
fs
b
1
X
5
pfc
b
1
X
6
con
stress
1
X
7
con
chronicstress
1
X
8
Wfsb,pfcb 1
X
9
Wpsb,pfcb 1
X
10
con
connection
1
X
11
W
con
connection
,
W
1
X
12
con
enzyme
1
X
13
W
con
enzyme
,
W
con
connection
,
W
1
X
14
con
mRNA
1
X
15
W
con
mRNA
,
W
con
enzyme
,
W
con
connection
,
W
1
X
16
con
DNA
1
X
17
Wcon
DNA
, Wcon
mRNA
, Wcon
enzyme
,Wcon
connection
,W1
X
18
con
miRNA
1
X
19
con
therapy
1
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
15
mcfp combi. param. algoistic Kolom1 Kolom2 Kolom22
function 1 2 1 2
parameter σ τ α β
X
1
ss
s
5 0.5
X
2
srs
s
5 0.5
X
3
ps
b
5 0.8
X
4
fs
b
5 0.6
X
5
pfc
b
5 0.5
X
6
con
stress
5 0.5
X
7
con
chronicstress
120 280
X
8
W
fsb,pfcb 5 0.5
X
9
Wpsb,pfcb
5 0.5
X
10
con
connection
5 0.5
X
11
W
con
connection
,
W
5 0.5
X
12
con
enzyme
5 0.5
X
13
Wcon
enzyme
,Wcon
connection
,W20 0.5
X
14
con
mRNA
5 0.5
X
15
Wcon
mRNA
, Wcon
enzyme
,Wcon
connection
,W20 0.5
X
16
con
DNA
5 0.5
X
17
Wcon
DNA
, Wcon
mRNA
, Wcon
enzyme
,Wcon
connection
,W20 0.7
X
18
con
miRNA
5 0.5
X
19
con
therapy
450 610
References
An, X., Guo, W., Wu, H., Fu, X., Li, M., Zhang, Y., Li, Y., Cui, R., Yang, W., Zhang, Z., &
Zhao, G. (2022). Sex Differences in depression caused by early life stress and related
mechanisms. Frontiers in Neuroscience, 16, Article 797755. https://doi.org/10.3389/
fnins.2022.797755
Bartlett, A. A., Singh, R., & Hunter, R. G. (2017). Anxiety and epigenetics.
Neuroepigenomics in Aging and Disease, 145–166.
David, F., Kalibala, G., Pichon, B., & Treur, J. (2023). A network model for modulating
sensory processing sensitivity in autism spectrum disorder: Epigenetics, adaptivity,
and other factors. In Proc. BICA*AI’23.
Fessler, D. M. T., Clark, J. A., & Clint, E. K. (2015). Evolutionary psychology and
evolutionary anthropology. In D. M. Buss (Ed.), The handbook of evolutionary
psychology (pp. 1029–1046). Wiley and Sons.
Fu, Z., Wang, L., Li, S., Chen, F., Au-Yeung, K. K. W., & Shi, C. (2021). MicroRNA as an
important target for anticancer drug development. Frontiers in Pharmacology, 12,
Article 736323.
Gold, A. L., Shechner, T., Farber, M. J., Spiro, C. N., Leibenluft, E., Pine, D. S., &
Britton, J. C. (2016). Amygdala–cortical connectivity: Associations with anxiety,
development, and threat. Depression and Anxiety, 33(10), 917–926.
Hendrikse, S. C. F., Treur, J., & Koole, S. L. (2023). Modeling emerging interpersonal
synchrony and its related adaptive short-term afiation and long-term bonding: A
second-order multi-adaptive neural agent model. International Journal of Neural
Systems, 33(7), 2350038 (41 pages).
Kendler, K. S., Zachar, P., & Craver, C. (2011). What kinds of things are psychiatric
disorders? Psychological Medicine, 41(6), 1143–1150. doi: 10.1017/
S0033291710001844.
Khan, S., & Khan, R. A. (2017). Chronic stress leads to anxiety and depression. Annals of
Psychiatry and Mental Health, 5(1), 1–4.
Kuranova, A., Booij, S. H., Menne-Lothmann, C., Decoster, J., van Winkel, R.,
Delespaul, P., De Hert, M., Derom, C., Thiery, E., Rutten, B. P. F., Jacobs, N., van
Os, J., Wigman, J. T. W., & Wichers, M. (2020). Measuring resilience prospectively as
the speed of affect recovery in daily life: A complex systems perspective on mental
health. BMC Medicine, 18(1), 36. https://doi.org/10.1186/s12916-020-1500-9
Lichtwarck-Aschoff, A., Kunnen, S. E., & van Geert, P. L. (2009). Here we go again: A
dynamic systems perspective on emotional rigidity across parent-adolescent
conicts. Developmental Psychology, 45(5), 1364–1375. https://doi.org/10.1037/
a0016713
Lin, E., & Tsai, S. J. (2020). Gene-environment interactions and role of epigenetics in
anxiety disorders. In Anxiety disorders: Rethinking and understanding recent discoveries
(pp. 93–102).
McEwen, B. S. (2017). Neurobiological and systemic effects of chronic stress. Chronic
Stress, 1, 2470547017692328.
Meaney, M. J. (2010). Epigenetics and the biological denition of gene ×environment
interactions. Child Development, 81(1), 41–79. https://doi.org/10.1111/j.1467-
8624.2009.01381.x
Miller, G. A., & Bartholomew, M. E. (2020). Challenges in the relationships between
psychological and biological phenomena in psychopathology. In J. Parnas, K. S.
Kendler, and P. Zachar (Eds.), Levels of analysis in psychopathology: Cross-disciplinary
perspectives (pp. 238–266). Cambridge University Press. doi: 10.1017/
9781108750349.022.
Mucha, M., Skrzypiec, A. E., Kolenchery, J. B., Brambilla, V., Patel, S., Labrador-
Ramos, A., … Pawlak, R. (2023). miR-483-5p offsets functional and behavioural
effects of stress in male mice through synapse-targeted repression of Pgap2 in the
basolateral amygdala. Nature Communications, 14(1), 2134.
Nelson, B., McGorry, P. D., Wichers, M., Wigman, J. T. W., & Hartmann, J. A. (2017).
Moving from static to dynamic models of the onset of mental disorder: A review.
JAMA Psychiatry, 74(5), 528–534. https://doi.org/10.1001/
jamapsychiatry.2017.0001
Nigg, J. T. (2017). Annual research review: On the relations among selfregulation, self-
control, executive functioning, effortful control, cognitive control, impulsivity, risk-
taking, and inhibition for developmental psychopathology. Journal of Child
Psychology and Psychiatry, 58(4), 361–383. https://doi.org/10.1111/jcpp.12675
Nigg, J. T. (2023). Considerations toward an epigenetic and common pathways theory of
mental disorder. Journal of Psychopathology and Clinical Science, 132(3), 297–313.
O’Donnell, K. J., & Meaney, M. J. (2020). Epigenetics, development, and
psychopathology. Annual Review of Clinical Psychology, 16, 327–350. doi: 10.1146/
annurev-clinpsy-050718-095530.
Pe˜
na, C. J., & Nestler, E. J. (2018). Progress in epigenetics of depression. Progress in
Molecular Biology and Translational Science, 157, 41–66.
Penner-Goeke, S., & Binder, E. B. (2019). Epigenetics and depression. Dialogues in Clinical
Neuroscience, 21(4), 397–405. https://doi.org/10.31887/DCNS.2019.21.4/ebinder
Rosa, J. M., Formolo, D. A., Yu, J., Lee, T. H., & Yau, S. Y. (2022). The role of MicroRNA
and microbiota in depression and anxiety. Frontiers in Behavioral Neuroscience, 16,
Article 828258.
Schiepek, G., Heinzel, S., Karch, S., Pl¨
oderl, M., & Strunk, G. (2016). Synergetics in
psychology: Patterns and pattern transitions in human change processes. In G.
Wunner, and A. Pelster (Eds.), Selforganization in complex systems: The past, present,
and future of synergetics (pp. 181–208). Cham: Springer. doi: 10.1007/978-3-319-
27635-9_12.
Schiepek, G. K., Tominschek, I., & Heinzel, S. (2014). Self-organization in psychotherapy:
Testing the synergetic model of change processes. Frontiers in Psychology, 5, 1089.
https://doi.org/10.3389/fpsyg.2014.01089
Scott, K. A., Hoban, A. E., Clarke, G., Moloney, G. M., Dinan, T. G., & Cryan, J. F. (2015).
Thinking small: Towards microRNA-based therapeutics for anxiety disorders. Expert
Opinion on Investigational Drugs, 24(4), 529–542.
S. Kathusing et al.
Cognitive Systems Research 83 (2024) 101177
16
Treur, J. (2016). Network-oriented modeling: Addressing complexity of cognitive, affective
and social interactions. Cham: Springer Nature.
Treur, J. (2020a). Network-oriented modeling for adaptive networks: Designing higher-order
adaptive biological, mental and social network models. Cham: Springer Nature.
Treur, J. (2019). Modeling higher-order adaptive evolutionary processes by multilevel
adaptive agent models. In: M. Baldoni, M. Dastani, B. Liao, Y. Sakurai, and R. Zalila
Wenkstern (Eds.), Principles and practice of multi-agent systems. Proceedings of PRIMA
2019. Lecture notes in computer science (Vol. 11873, pp. 505–513). Cham: Springer
Nature. doi: 10.1007/978-3-030-33792-6_35.
Treur, J. (2020b). Modeling higher-order adaptive evolutionary processes by reied
adaptive network models. In J. Treur, Network-oriented modeling for adaptive
networks: Designing higher-order adaptive biological, mental and social network models
(Ch 7, pp. 167–185).
Treur, J. (2021). On the dynamics and adaptivity of mental processes: relating adaptive
dynamical systems and self-modeling network models by mathematical analysis.
Cognitive Systems Research, 70, 93–100.
Wichers, M., Wigman, J. T. W., & Myin-Germeys, I. (2015). Micro-level affect dynamics
in psychopathology viewed from complex dynamical system theory. Emotion Review,
7(4), 362–367. https://doi.org/10.1177/1754073915590623
S. Kathusing et al.
