Hybrid Petri Nets

for Modelling the Eukaryotic Cell Cycle

Mostafa Herajy1 , Martin Schwarick2, and Monika Heiner2
1

Port Said University, Faculty of Science,

Department of Mathematics and Computer Science,
42521 - Port Said, Egypt
2
Brandenburg University of Technology at Cottbus,
Computer Science Institute,
Data Structures and Software Dependability,
Postbox 10 13 44, 03044 Cottbus, Germany
http://www-dssz.informatik.tu-cottbus.de/

Abstract. System level understanding of the repetitive cycle of cell

growth and division is crucial for disclosing many unknown principles
of biological organisms. The deterministic or stochastic approach when
deployed separately are not sucient to study such cell regulation due
to the complexity of the reaction network and the existence of reactions
at dierent time scales. Thus, an integration of both approaches is advisable to study such biochemical networks. In this paper we show how
Generalised Hybrid Petri Nets can be used to intuitively represent and
simulate the eukaryotic cell cycle. Our model captures intrinsic as well as
extrinsic noise and deploys stochastic as well as deterministic reactions.
Additionally, marking-dependent arc weights are biologically motivated
and introduced to Snoopy a tool for animating and simulating Petri
nets in various paradigms.
Keywords: Generalised hybrid Petri nets, hybrid modelling, eukaryotic
cell cycle, Snoopy, marking-dependent arc weight.

Introduction

The reproduction of eukaryotic cells is controlled by a complex regulatory network of reactions known as cell cycle [19,20,24]. During a cell cycle, cells grow,
replicate and divide into two daughter cells [13,21]. This regulation cycle consists
of four phases: S phase (synthesis) and M phase (mitosis) separated by two gap
phases: G1 and G2 [24]. During the S phase, the cell replicates all of its components, while it divides each component more or less evenly between the two
daughter cells at the end of the M phase [13]. After the S phase, there is a gap
(G2) where the cell ensures that the duplication of DNA has been completed
and prepares itself for mitosis. Newborn cells are not immediately replicated,
instead they are located at the G1 gap. The processes of synthesis and mitosis
alternate during the reproduction process; see Figure 1 for a graphical illustration of the cell cycle regulation process. Please note that the phases G1, S, and
M. Koutny et al. (Eds.): ToPNoC VIII, LNCS 8100, pp. 123141, 2013.
c Springer-Verlag Berlin Heidelberg 2013


124

M. Herajy, M. Schwarick, and M. Heiner

Fig. 1. Graphical illustration of the cell cycle [27]. The cell cycle consists of four distinct
phases: G1 , synthesis (S), G2 , and mitosis (M), respectively. The rst three phases are
known as interphase (referred to by the outer ring). Cells that have stopped dividing
enter the G0 phase.

G2 are commonly subsumed as interphase as indicated by the outer cycle in that

gure. Understanding such control cycles is crucial for revealing defects in cell
growth that underlies many human diseases (e.g., cancer) [25].
In the eukaryotic cell cycle, the alternation between the S and the M phase
as well as the balance of growth and division is governed by the activity of a
family of cyclin-dependent protein kinases (CDK) [24]. Therefore, many computational models have been constructed to study the control system of CDK
(e.g., in [1,13,19,20,24]). Some of these models are based on the deterministic approach which represents changes of species concentrations as continuous variables which evolve deterministically and continuously with respect to
time (in the following called continuous simulation). However, an important
requirement of the cell cycle model is to capture the variability of the cellular volume to reproduce the in vivo experiment results. Unfortunately, the
deterministic approach cannot capture such cellular volume variability [20]. Motivated by this argument, a number of stochastic models have been created and
simulated using either a stochastic simulation algorithm (e.g., [13]) or by introducing noise to the model through Langevin equation [22]. However, the stochastic approach is computationally expensive, particularly when the model under
study contains reactions with high rates and/or species with large numbers of
molecules.
The eukaryotic cell cycle model does indeed exhibit high rates of some reactions, while some other reactions have low rates, which are responsible for the

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

125

intrinsic noise due to molecular uctuations [13]. Similarly, the model contains
some species with a large number of molecules, while some other species have a
few number of molecules. The existence of reactions at dierent time scales (fast
and slow) suggests a simulation using a hybrid approach.
Generalised Hybrid Petri Nets (GHPN bio ) have been introduced in [10],
[11] and [12], to represent and simulate sti biochemical networks where fast
reactions are represented and simulated continuously, while slow reactions are
carried out stochastically. GHPN bio provide rich modelling and simulation functionalities by combining all features of Continuous Petri Nets [3] and Extended
Stochastic Petri Nets [16], including three types of deterministic transitions.
Moreover, the partitioning of reaction networks (i.e., the assignment of each
reaction to either the stochastic or the continuous paradigm) can either be
done o-line (statically, i.e., before the simulation starts) or on-line (dynamically, i.e., while the simulation is in progress). The implementation of GHPN bio
is available as part of Snoopy [7] - a tool to design and animate or simulate hierarchical graphs, among them qualitative, stochastic, continuous and
hybrid Petri nets. Indeed, the cell cycle model turns out to be an ideal case
study where the majority of the GHPN bio features can be demonstrated. Moreover, it makes a strong case for the introduction of marking-dependent arc
weights.
Another hybrid net class which provides functionalities related to GHPN bio is
known as Hybrid Functional Petri nets (HFPN) [18]. However, HFPN have been
developed to focus on hybrid (discrete/continuous) model construction where
stochastic transitions are not required. Moreover, modelling features like logical
nodes, hierarchy, and modier arcs, which are imperative when considering larger
models, are not supported [7].
In this paper we present another argument to motivate hybrid simulation of
the cell cycle control system. The cell cycle model contains some reactions which
would be better represented as continuous processes, specically the growth of
the cellular volume needs to be treated continuously, while other reactions of
low rates have to be considered as stochastic processes. For instance, Mura and
Csikasz-Nagy constructed in [19] a stochastic version of the model in [1] using stochastic Petri nets. However, they could not intuitively represent the cell
growth process which evolves continuously and exponentially with respect to
time using stochastic Petri net primitives only. Indeed, cell growth is a typical
example where continuous transitions are an appropriate means.
This paper is organised as follows: we start o by pinpointing some related
work. After that, a brief introduction of Generalised Hybrid Petri Nets is presented. To conveniently model the cell cycle regulation behaviour, we extend the
formal denition of GHPN bio , as they have been introduced in [10], to include
marking-dependent arc weights. Next, we discuss a hybrid Petri net model of the
eukaryotic cell cycle and discuss in detail some of its key modelling components.
In Section 5 we show the simulation results produced by Snoopys hybrid simulation engine and compare them to the continuous and stochastic ones. Finally,
we sum up with conclusions and outlook.

126

M. Herajy, M. Schwarick, and M. Heiner

Related Work

Mura and Csikasz-Nagy converted the deterministic model of Chen et al. [1]
into a stochastic Petri net [19] to study the eect of noise on cell cycle progression. However, some components could not intuitively be modelled using
stochastic Petri net primitives only (e.g., cell growth). Moreover, their model
is based on phenomenological rate laws (e.g., Michaelis-Menten) which do not
work well with stochastic simulation algorithms [13]. Sabouri-Ghomi et al. [20],
and Kar et al. [13] asserted that applying Gillespies stochastic simulation algorithm [4,5] directly to phenomenological rate laws might produce incorrect
results. Therefore, they unpacked the deterministic model of Tyson-Novak [24]
(who use non-elementary reaction kinetics, e.g., Michaelis Menten and Hill functions) to express it completely in terms of elementary mass-action kinetics. The
Tyson-Novak model is based on a bistable switch between the complex CycBCdk1 (denoted by variable X) and the complex Cdh1-APC (denoted by the variable Y). CycB-Cdk1 phosphorylates Cdh1-APC and free Cdh1-APC catalyses
the degradation of CycB-Cdk1. Figure 2 presents a continuous Petri net representation of the Tyson-Novak model. To model a complete cell cycle, Kar et al.
[13] unpacked the eect of Cdc20 and Cdc14 which are lumped in the variable Z
in the Tyson-Novak model. High activity of CycB-Cdk1 promotes the synthesis
of Cdc20 which activates Cdc14. Finally the dephosphorylated Cdc14 activates
Cdh1-APC. The Kar et al. model accounts for both intrinsic and extrinsic noise.
Intrinsic noise is due to the uctuation of species with low numbers of molecules,
while extrinsic noise is due to the unequal division of the cell between the two
daughter cells [13].
In [2] and [17], two detailed HFPN models are constructed for the Fission yeast
and Xenopus cell cycles, respectively. However, intrinsic noise, which is necessary
for reproducing the variability of the cellular volume, is not captured because HFPN
do not support the (full) interplay between stochastic and continuous regimes.
Thus, these models are built using the hybrid (discrete/continuous) paradigm.
In [21], a hybrid model, which combines ordinary dierential equations (ODEs)
and discrete boolean networks, has been constructed to integrate quantitative as
well as qualitative parts in one model. The latter approach requires less knowledge of realistic kinetic rate constants. Liu et al. [15] simulate the stochastic
model of [13] using the Haseltine and Rawlings approach [6]. However, such
models cannot be represented structurally or graphically which makes their
maintenance and extension more diecult.
In this paper a hybrid Petri net model of the eukaryotic cell cycle is presented
as a sophisticated example for the kind of hybrid models that can be constructed
using GHPN bio . The model is hybrid in the sense that it combines continuous,
stochastic and immediate transitions to represent deterministic, stochastic and
control behaviour. Our main goal is to show how such a class of models is intuitively represented and executed using hybrid Petri net primitives. Besides, Petri
nets analysis tools can be applied to the constructed models as well [8]. Using
Snoopys simulator, cell cycle models incorporating continuous net components
can be simulated using either the continuous or hybrid engine.

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

127

V
10
X
6

Yp

Y
2189

Fig. 2. A continuous Petri net representation of the Tyson-Novak model [24]: X (CycBCdk1) phosphorylates Y (Cdh1-APC) and free Y catalyses the degradation of X. Z
denotes the eects of Cdc20 and Cdc14. High activity of X promotes the synthesis of
Cdc20 which activates Cdc14. The dephosphorylated Cdc14 activates Y. This behaviour
results in a bistable switch that is responsible for the transitions between G1 and
S-G2-M states.

Generalised Hybrid Petri Nets

To model sti biochemical networks, GHPN bio [10] combine both stochastic
and continuous elements in one and the same model. Indeed, continuous and
stochastic Petri nets complement each other. Fluctuation and discreteness can
conveniently be modelled and simulated in the stochastic paradigm and at the
same time, the computational expensive parts can be simulated deterministically
via ODE solvers. Modelling and ecient simulation of sti biochemical networks
(i.e., networks that contain reactions at more than one time scale) are helpful
functionalities that GHPN bio provide for systems biology.
Generally speaking, biochemical systems can involve reactions from more than
one type of biological networks, for instance gene regulation, metabolic pathways
or signal transduction pathways. Incorporating reactions which belong to distinct
(biological) network types, tends to result into sti systems. This follows from
the fact that, e.g., species in gene regulation networks may contain few numbers
of molecules, while species in metabolic networks often contain large numbers of
molecules [14].
In the rest of this section, we will give a brief introduction of GHPN bio in
terms of the graphical representation of its elements as well as the ring rule
and connectivity between the continuous and stochastic net parts. The formal
semantics is given in [10].

128

3.1

M. Herajy, M. Schwarick, and M. Heiner

Elements

The GHPN bio elements are classied into three categories: places, transitions,
and arcs.
GHPN bio oer two types of places: discrete and continuous. A discrete place
(single line circle) holds a non-negative integer number which represents, e.g.,
the number of molecules of a given species (tokens in Petri net notions). A
continuous place (shaded line circle) holds a non-negative real number which
represents, e.g., the concentration of a given species.
Furthermore, GHPN bio oer ve transition types: stochastic, immediate, deterministically delayed, scheduled, and continuous transitions [8]. Stochastic
transitions, which are drawn in Snoopy as a square, re with an exponentially
distributed random delay. The user can specify a set of ring rate functions,
which determine the random ring delay. The transitions pre-places can be
used to dene the ring rate functions of stochastic transitions. Immediate transitions (black bar) re with zero delay, and have always highest priority to re.
They may carry weights which specify the relative ring frequency in the case
of conicts between immediate transitions. Deterministically delayed transitions
(black square) re after a specied constant time delay. Scheduled transitions
(grey square) re at user-specied absolute time points. Continuous transitions
(shaded line square) re continuously in the same way like in continuous Petri
nets. Their semantics is governed by ODEs which dene the continuous change
in the transitions pre- and post-places. More details about the biochemical interpretation of deterministically delayed, scheduled, and immediate transitions
can be found in [9] and [16]. To simplify the presentation, we occasionally refer
to stochastic, immediate, deterministically delayed or scheduled transitions as
discrete transitions.
The connection between those two types of nodes (places and transitions)
takes place using a set of dierent arcs (edges). GHPN bio oer six types of arcs:
standard, inhibitor, read, equal, reset and modier arcs. Standard arcs connect
transitions with places or vice versa. They can be discrete, i.e., carry non-negative
integer-valued weights (stoichiometry in the biochemical context), or continuous,
i.e., carry non-negative real-valued weights. In addition to their inuence on the
enabling of transitions, they also aect the place marking when a transition res
by removing (adding) tokens from (to) the transitions pre-places (post-places).
Extended arcs like inhibitor, read, equal, reset, and modier arcs can only
be used to connect places with transitions, but not vice versa. A transition
connected with an inhibitor arc is enabled (with respect to this pre-place) if
the marking of the pre-place is less than the arc weight. Contrary, a transition
connected with a read arc is enabled if the marking of the pre-place is greater
than or equal to the arc weight. Similarly, a transition connected using an equal
arc is enabled if the marking of the pre-place is equal to the arc weight.
The other two remaining arcs do not aect the enabling of transitions. A reset
arc is used to reset a place marking to zero when the corresponding transition
res. Modier arcs permit to include any place in the transitions rate functions
and simultaneously preserve the net structure restriction.

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

129

Places
Discrete

Continuous

Transitions

<1>
Stochastic

Continuous

[_SimStart,1,_SimEnd]

Immediate Deterministic

Scheduled

Arcs

Standard

Read

Inhibitor

Equal

Reset

Modifier

Fig. 3. Graphical representation of the GHPN bio elements. Places are classied as
discrete and continuous, transitions as stochastic, continuous, immediate, deterministically delayed, and scheduled, and arcs as standard, inhibitor, read, equal, reset, and
modier.
p2

p2

p1/2
p1
p1

<1>

(a)

p3

p1

p3

<1>
p1/2

(b)

Fig. 4. Marking-dependent arc weights illustrated by a simple biological example. (a)

cell division cannot be modelled, (b) cell division can intuitively be modelled. The numbers between angle brackets are the delays of the deterministically delayed transitions.
Later we will assume that cell division does not consume time.

The connection rules and their underlying formal semantics are discussed in
more details below. Figure 3 provides a graphical illustration of all elements. Although this graphical notation is the default one, it can easily be customised
using Snoopy, the Petri nets editing tool. To support special modelling requirements of some biological models (e.g., the cell cycle model), we extended
GHPN bio to permit pre-places of a transition as arc weight, similar to the idea

130

M. Herajy, M. Schwarick, and M. Heiner

of self-modifying nets which has been originally introduced in [26], or even a

function which is dened in terms of a transitions pre-places [18].
Consider the following simple biological example. When a cell divides its mass
between two daughter cells, each daughter obtains approximately half of the
mass. This example cannot easily be modelled using discrete Petri nets. Moreover, there is no way to model it if the mass is represented by a continuous place
as shown in Figure 4a. In Figure 4b, using marking-dependent arc weights; the
ingoing arc of the transition t has a weight equal to the marking of the place p1 ,
while each of the two outgoing arcs has a weight equal to half of the marking of
that place.
Motivated by the case study discussed in this paper, marking-dependent arc
weights have been introduced for the majority of arc types supported by Snoopy
(standard, read, inhibitor, and equal arc). For more details see Section 4.2.
3.2

Connection Rules

An important question arises when considering the combination of discrete and

continuous elements: how are these two dierent parts connected with each
other? Figure 5 provides a graphical illustration of how the connection between
dierent elements of GHPN bio takes place.
First, we will consider the connection between continuous transitions and
the other elements of GHPN bio . Continuous transitions can be connected with
continuous places in both directions using continuous arcs (i.e., arcs with realvalued weight). This means that continuous places can be pre- or post-places
of continuous transitions. These connections typically represent deterministic
biological interactions.
Continuous transitions can also be connected with discrete places, but only by
one of the extended arcs (inhibitor, read, equal, and modier). This type of connection allows a link between discrete and continuous parts of a biochemical model.
Discrete places are not allowed to be connected with continuous transitions
using standard arcs, because the ring of continuous transitions is governed by
ODEs which require real values in the pre- and post-places. Hence, this cannot
take place in the discrete world.
Second, discrete transitions can be connected with discrete or continuous
places in both directions using standard arcs. However, the arc weight needs to
be considered. The connection between discrete transitions and discrete places
takes place using arcs with non-negative integer numbers, while the connection
between continuous places and discrete transitions is weighted by non-negative
real numbers. The general rule to determine the weight type of arcs is to follow
the type of the connected place.
3.3

Formal Definition

In this section, the syntax of GHPN bio is formally dened to include the makingdependent arc weight. The formal semantics including the enabling and ring
rules as well as the conict resolution are given in [10].

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

131

or

or

or

or

or

or

or

or

Discrete Transition

Continuous Transition

Fig. 5. Possible connections between GHPN bio elements. The restrictions are: discrete
places cannot be connected with continuous transitions using standard arcs, continuous
places cannot be tested with equal arcs, and continuous transitions cannot use reset
arcs.

Definition 1 (Generalised Hybrid Petri Nets). A Generalised Hybrid Petri

Net is a 6-tuple GHPN bio = [P, T, A, F, V, m0 ], where P , T are finite, non-empty
and disjoint sets. P is the set of places, and T is the set of transitions with:
P = Pdisc Pcont whereby Pdisc is the set of discrete places to which nonnegative integer values are assigned, and Pcont is the set of continuous places
to which non-negative real values are assigned.
T = TD Tcont ,
TD = Tstoch Tim Ttimed Tscheduled with:
1. Tstoch is the set of stochastic transitions, which fire randomly after exponentially distributed waiting time.
2. Tim is the set of immediate transitions, which fire with waiting time zero;
they have higher priority compared with other transitions.
3. Ttimed is the set of deterministically delayed transitions, which fire after
a deterministic time delay.
4. Tscheduled is the set of scheduled transitions, which fire at predefined time
points.
5. Tcont is the set of continuous transitions, which fire continuously over
time.
A = Adisc Acont Ainhibit Aread Aequal Areset Amodif ier is the set
of directed arcs, with:
1. Adisc ((P T ) (T P )) defines the set of discrete arcs.

132

M. Herajy, M. Schwarick, and M. Heiner

2.
3.
4.
5.
6.
7.
the

Acont ((Pcont T ) (T Pcont )) defines the set of continuous arcs.

Aread (P T ) defines the set of read arcs.
Ainhibit (P T ) defines the set of inhibits arcs.
Aequal (Pdisc T ) defines the set of equal arcs.
Areset (P TD ) defines the set of reset arcs,
Amodif ier (P T ) defines the set of modifier arcs.
function F

Acont Dq ,

Adisc Dn ,

Dq ,
A

read
F : Ainhibit Dq ,

Aequal Dn ,

A
{1},

reset

Amodif ier {1}.

assigns a marking-dependent function to each arc, where Dn and Dq are sets

of functions defined as follows:
| tj |

Dn = {dn |dn : N0

| tj |

Dq = {dq |dq : R0

N, tj T },

Q+ , tj T }.

V is a set of functions V = {g, d, w, f } where :

1. g : Tstoch Hs is a function which assigns a stochastic hazard function
| t |
hst to each transition tj Tstoch , whereby Hs = {hst |hst : R0 j
R+
0 , tj Tstoch } is the set of all stochastic hazard functions, and g(tj ) =
hst , tj Tstoch .
2. w : Tim Hw is a function which assigns a weight function hw to each
| t |

immediate transition tj Tim , such that Hw = {hwt |hwt : R0 j

R+
0 , tj Tim } is the set of all weight functions, and w(tj ) = hwt , tj
Tim .
3. d : Ttimed Tscheduled R+
0 , is a function which assigns a constant time
to each deterministically delayed and scheduled transition representing
the (relative or absolute) waiting time.
4. f : Tcont Hc is a function which assigns a rate function hc to each
| t |
continuous transition tj Tcont , such that Hc = {hct |hct : R0 j
R+
0 , tj Tcont } is the set of all rates functions and f (tj ) = hct , tj
Tcont .

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

133

m0 = mdisc mcont is the initial marking for both the continuous and discrete
|Pcont |

places, whereby mcont R0

|Pdisc |

, mdisc N0

Here, N denotes the set of natural numbers excluding 0, N0 denotes the set of nonnegative integer numbers, R0 denotes the set of non-negative real numbers, Q+ denotes the set of positive rational numbers, and tj denotes the set of pre-places of a
transition tj .

A distinguishing feature of GHPN bio compared with other hybrid Petri net
classes is its support of the full interplay between stochastic and continuous
transitions. Such interplay is implemented by updating and monitoring the rates
of stochastic transitions while numerically solving the set of ODEs induced by
the continuous transitions (For more details see [10]). By this way, accurate
results are obtained during simulation.

The Model

Figure 6 shows the hybrid Petri net model which has been developed based on
the previous one introduced by Kar et al. in [13]. Proteins, genes, and mRNAs
are represented by places, reactions by transitions. We use the same kinetic
parameters and initial values as in [13]. For the sake of space we do not repeat
the kinetic parameters, but the initial marking is shown on the places. Moreover,
we use Snoopys logical node feature to simplify the connections between nodes.
For example, place X and Y are involved in many reactions which decreases
the networks readability. We repeat those nodes multiple times with the same
names to keep the model understandable (logical places). Likewise, the transition
divide is a logical transition. Furthermore, the increase of the cellular volume is
intuitively represented using a continuous transition with a rate V , where is
the growth factor and V is the cellular volume, modelled as a continuous place.
The model contains three dierent transition types: continuous, stochastic,
and immediate. Continuous transitions simulate the corresponding reactions deterministically, while stochastic transitions carry them out stochastically. The
latter transitions are responsible for molecular uctuations. Immediate transitions monitor the model evolution and perform the division when the free number
of molecules of Cdh1 APC reaches a certain threshold (Y = Y + Y X + X Y ).
In the sequel we discuss in more detail some of the models key components
and the corresponding GHPN bio representations.
4.1

Decision to Perform Division

In this section we consider the process of division in more detail. When the
number of molecules of Y becomes greater than a certain threshold (in our case

134

M. Herajy, M. Schwarick, and M. Heiner




*


6




#%





#)



#*

6



#%



#



#

*

!
!

#

"

#

"

#



"

"


-






+##

#


&

!

"6



!

$


$"

"6

'


$


$6

#%
"

#'

$"6

'



##

$"

6

"



)



#


78 9


78 

:
$78 &8;8 *











6


&


6

'!!


6


#

% 









 

*'

*%

*)

6

 
(

"


(

!

"


& "

"6






"
!









6

#!

6



6
%
'



(6

)

*#



#









*



6

)



**



%



#

'

#

#%

!

"6
"



Fig. 6. A GHPN bio representation of the eukaryotic cell cycle. The model employs
dierent types of transitions: continuous, stochastic and immediate. All reactions affecting mRNAs are represented and simulated stochastically. Repetitive nodes (places
and transitions) with same names are logical nodes. When the immediate transition
divide res, it divides the current place marking more or less equally. Equal division means that the cellular volume of the daughter cell is always half of its parent.
This model could be easily extended to permit unequal division, where a random
variation in the cellular volume is possible, by having arc weights with random functions. The unequal division type will reproduce extrinsic noise. The type of division
(equal, or unequal) depends on the outgoing arc weight and its eect is implemented by
marking-dependent arc weights.

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

135

1200), the cell divides the cellular volume and other components (e.g., mRNAs)
between the two daughter cells. In Figure 7a, this process is represented by the
immediate transition check with a weight dened by the Boolean expression
Y > threshold (the weight is 0 if the Boolean expression yields false, and 1
for the result true). Recall that weights of immediate transitions determine the
ring frequencies of immediate transitions in the case of conicts. A weight of
zero means that a transition cannot re at all. However, when the transition
check has a weight of one, it adds a token to the place ready to divide which
triggers the transition divide to carry out the division. To give the transition
divide a chance to re before re-checking the value of Y , an inhibitor arc is used
as constraint. Please note that the transitions critical and check need the current
marking of the places X Y , Y , and Y X only to calculate the term Y in the
transitions weight. Therefore, modier arcs are used to full this requirement.
An interesting characteristics of the model is the division process. Although
the division can take place when the value of Y is greater than a certain threshold, it does not do that all the times. For example, at the beginning of the
simulation, the initial value of Y satises the division criteria. However; the
cell should not divide because it is still at G1 phase which means that it has
to replicate itself before it can divide. We model these cases by adding a new
immediate transition which detects the critical value of Y , before checking for
division. Therefore the transition critical monitors the value of Y . When the
value of Y goes below a certain threshold, it enables the division process.
4.2

Cell Division and Marking-Dependent Arc Weights

When a cell divides, it splits all of its components more or less evenly between
the two daughter cells. This is most naturally expressed with marking-dependent
arc weights [26]. In Figure 7a, when the transition divide res, it removes all of the
current marking of the place V and adds V/2 to it. To permit uneven division of the
cell volume and other components, arc weights can be a function which operates on
the current place marking [18]. However, we restrict the places used in arc weights
to a transitions pre-places to keep the locality principle Petri nets are famous for.
Figure 7b illustrates the process of cell division graphically by showing a
simulation trace.
Moreover, all proteins and mRNAs have to undergo such division. This means
the transition divide has to be connected with each place in the net that represents a protein or mRNA. The ingoing arc weight of such a connection is equal
to the pre-places current marking, while the outgoing arc weight is equal to half
of the pre-places current marking. Furthermore, the markings of discrete places
are rounded after the division process to preserve the discrete representation of
the molecular species.
4.3

Transition Partitioning

The model in Figure 6 contains transitions which re at dierent rates. For

instance, transition R3 res more frequently than R1 as illustrated in Figure

136

M. Herajy, M. Schwarick, and M. Heiner

ready_for_check

Y: Cdh1APC
V: Cellular Volume

X_Y
36

Y
divide

2189
critical

Y_X

check

ready_for_divide
V

V/2

83

21
V

(a)

(b)

Fig. 7. Cell Division (a) A sub-net for modelling the decision of the division process
(see also upper right corner of Figure 6). The transition critical monitors the value
of Y and adds a token to ready for check when Y < 300. Later, when the value of
Y increases and becomes greater than a certain threshold (1200), the transition check
res and adds a token to ready for divide which signals the transition divide to perform
the division. Inhibitor arcs are used as checkpoints for the sequence of events: critical
check divide. (b) Hybrid simulation trace of cell division.

8a. Slow transitions should be simulated stochastically to account for molecular

uctuations, while fast transitions need to be simulated continuously for the
sake of numerical eciency. Indeed, transitions of the latter type consume the
majority of computational resources.
In this model, transitions are statically partitioned before the simulation
starts. The transition type is determined by executing a single run and analysing
the results as shown in Figure 8. Increasing (decreasing) the accuracy of the simulation results involves converting more continuous (stochastic) transitions into
stochastic (continuous) ones.
Another approach for partitioning is to perform it dynamically during the simulation. Using this technique, a transition changes its type from stochastic to continuous or vice versa according to the current ring rate. GHPN bio provide the
user with a trade-o between eciency and accuracy by permitting the user to
specify two thresholds: a0min and a0max , the minimum and maximum cumulative
propensity (i.e., the total rates of stochastic transitions), respectively. Moreover,
two other thresholds are required to perform dynamic partitioning: the place marking threshold and the transition rate threshold. The former is used to ensure that
species concentrations are large enough to be simulated continuously, while the
latter is used to partition transitions into fast and slow based on their rates. A
transition is simulated continuously, if its rate exceeds the rate threshold and the
marking of all its pre-place is greater than the marking threshold.
Nevertheless, cell growth has to be represented and simulated continuously
in both partitioning approaches. Using o-line partitioning, this can be easily
communicated to the simulator by drawing a continuous transition. However, in
the case of dynamic partitioning, the transition rate threshold had to be set less
than the smallest expected rate of cell growth which makes the latter approach
unsuitable for the cell cycle model.

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

137

6000
R1
R3

R26
R18

8000
5000
7000

6000

4000

Rates

Rates

5000

4000

3000

3000

2000

2000
1000
1000

0
0

100

200

300

400

500
time

600

700

(a)

800

900

1000

100

200

300

400

500
time

600

700

800

900

1000

(b)

Fig. 8. Example of dierent transition ring rates. (a) transition R3 res more frequently than the transition R1 , and (b) transition R18 res much more often than R26 .

Simulation Results

In this section we show some simulation results of the model in Figure 6 using
Snoopys hybrid simulator. Figures 9 - 12 present time course simulation results
of some model species of continuous and hybrid trajectories.
In the hybrid setting, species of low numbers of molecules are simulated using the stochastic regime, e.g., mRNAx and mRNAz ; thus, their numbers of
molecules show variability. Such variability is due to the intrinsic noise which is
captured by the stochastic simulation algorithm.
Figure 12 compares continuous and hybrid simulation results of the cellular
volume (V ). Using continuous simulation, parent cells divide all the time equally,
and the model does not produce variability in its volume size. Contrary, hybrid
simulation does show variability in the cellular volume because species of low
numbers of molecules (e.g., mRNAs) are simulated stochastically.
The variability behaviour in the cellular volume, which is produced by the
hybrid simulation, is close to the biological model behaviour. For example, the
Fission yeast cells have at division a Coecient of Variation (CV) of the cellular
volume of about 6% [13]. The CV is a normalised measure of dispersion of a
probability distribution. It is used to judge the variability of a result and it is
dened as the ratio of the standard deviation to the mean , i.e, CV = .
Table 1 compares the CV and mean values of the deterministic, stochastic,
and hybrid simulation results as well as the experimental data of the Fission
yeast (wild-type). The continuous and hybrid results are computed by exporting
the Snoopy simulation output to a comma-separated values format (CSV). Then
a tiny script extracts the dierent statistics, i.e., and CV.
As expected, the CVs of continuous simulation results are zero. This means
that continuous simulation does not exhibit any variability in the cellular volume.
Moreover, the stochastic and hybrid statistics are similar, but not the same. The
variability of cellular volumes of cells simulated via the hybrid version is slightly
less than the corresponding stochastic simulation. However, this is an expected
behaviour since some of the transitions are continuously simulated.

138

M. Herajy, M. Schwarick, and M. Heiner

Y hybrid
2000

1500

1500
concentration

concentration

Y continuous
2000

1000

500

1000

500

0
0

100

200

300

400

500
time

600

700

800

900

1000

100

200

300

(a)

400

500
time

600

700

800

900

1000

(b)

Fig. 9. Time course result of Y; (a) continuous, and (b) hybrid

12

12
mRNAx continuous

mRNAx hybrid

8
concentration

10

concentration

10

0
0

100

200

300

400

500

600

700

800

900

1000

100

200

300

400

time

500

600

700

800

900

1000

900

1000

time

(a)

(b)

Fig. 10. Time course result of mRNAx; (a) continuous, and (b) hybrid
6

6
mRNAz hybrid

4
concentration

concentration

mRNAz continuous

0
0

100

200

300

400

500
time

(a)

600

700

800

900

1000

100

200

300

400

500

600

700

800

time

(b)

Fig. 11. Time course result of mRNAz; (a) continuos, and (b) hybrid

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

35

35

30

139

30

25

25

20
size

size

20

15

15

10

10

0
500

600

700

800

900

1000

0
500

600

700

time

800

900

1000

time

(a)

(b)

Fig. 12. Time course results of the cellular volume (V); (a) continuous, and (b) hybrid
simulation
Table 1. Comparison of the continuous, stochastic, and hybrid simulation results of
the model in Figure 6. The volume size is given in (femtolitre).

cell age, min size at division, size at birth,

reference
mean CV% mean CV %
mean CV%

No.

simulator

Fission yeast 148

10.8

14.4

5.9

8.2

6.3

deterministic 115.9

30.9

15.9

3
4

stochastic
hybrid

13
12

29.1
29.9

8.2
7.4

14.5
15

8.2
7.4

[13]
-

115.5
115.5

[23]

Conclusions and Outlook

In this paper we have shown a class of biological models that can appropriately be
modelled using hybrid Petri nets. As an example we have presented and discussed
a hybrid Petri net model of the eukaryotic cell cycle. This specic model can be
executed using either continuous or hybrid simulators. It employs continuous,
stochastic and immediate transitions to intuitively represent the entire model
logic. Generally, depending on the type of model, a GHPN bio model can be
simulated continuously, stochastically or in a hybrid way.
The model is implemented using Snoopy. The model itself and the tool are
available at http://www-dssz.informatik.tu-cottbus.de/. Marking-dependent arc weights are a new feature recently added to Snoopy which is currently
not available in the ocial Snoopy release. However, the under-development
version is freely available on request.
Comparing the simulation results we notice that hybrid simulation produces
results close to the stochastic ones (in terms of the resulting CVs), while simulation eciency could be preserved. Indeed, the reactions of this model can
easily be split into slow and fast reactions, which makes it an ideal case study
for hybrid simulation algorithms.

140

M. Herajy, M. Schwarick, and M. Heiner

Marking-dependent arc weights are of paramount importance to model such

biological scenarios since they provide a direct tool to program certain biological
phenomenon (e.g., cell division). Therefore, we intend to add even more functionalities into this direction to permit more user-dened operators depending
on the transitions pre-places, e.g., random function.
So far the partitioning of the reactions into stochastic and deterministic ones
is carried out using a heuristic approach (see Section 4.3). However, (as suggested
by one of the reviewers) a more sophisticated partitioning could be performed. For
instance, the fast processes could be described by a quasi (or pseudo) steady state
approach, assuming that they reach equilibrium rapidly. In other words, they could
be better described by setting the corresponding ODEs to zero and solving them.
In contrast, continuous dynamics could be seen as more appropriate for abundant molecules whose concentration display a small coecient of variation, and
stochastic dynamics for those molecules evolving at low copy numbers.
Finally, the model presented in this paper could be viewed as a sub-net in a
bigger network of reactions (e.g., modelling budding yeast cell cycle or Fission
yeast cells). Snoopys hierarchical nodes might simplify such task as they provide
an easy tool to insert a sub-net into a bigger one.
Acknowledgements. This work was done during the stay of Mostafa Herajy
at the Brandenburg University of Technology at Cottbus, Germany, supported
by the GERLS (German Egyptian Research Long Term Scholarships) program,
which is administered by the DAAD in close cooperation with the MHESR and
German universities.

References
1. Chen, K., Calzone, L., Csikasz-Nagy, A., Cross, F., Novak, B., Tyson, J.: Integrative
analysis of cell cycle control in budding yeast. Mol. Biol. Cel. 5(8), 38413862 (2004)
2. Fujita, S., Matsui, M., Matsuno, H., Miyano, S.: Modeling and simulation of ssion
yeast cell cycle on hybrid functional Petri net. IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences E87-A(11), 29192927
(2004)
3. Gilbert, D., Heiner, M.: From Petri nets to dierential equations - an integrative
approach for biochemical network analysis. In: Donatelli, S., Thiagarajan, P.S.
(eds.) ICATPN 2006. LNCS, vol. 4024, pp. 181200. Springer, Heidelberg (2006)
4. Gillespie, D.: A general method for numerically simulating the stochastic time
evolution of coupled chemical reactions. J. Comput. Phys. 22(4), 403434 (1976)
5. Gillespie, D.: Stochastic simulation of chemical kinetics. Annual Review of Physical
Chemistry 58(1), 3555 (2007)
6. Haseltine, E., Rawlings, J.: Approximate simulation of coupled fast and slow reactions for stochastic chemical kinetics. J. Chem. Phys. 117(15), 69596969 (2002)
7. Heiner, M., Herajy, M., Liu, F., Rohr, C., Schwarick, M.: Snoopy a unifying Petri
net tool. In: Haddad, S., Pomello, L. (eds.) PETRI NETS 2012. LNCS, vol. 7347,
pp. 398407. Springer, Heidelberg (2012)
8. Heiner, M., Gilbert, D., Donaldson, R.: Petri nets for systems and synthetic biology.
In: Bernardo, M., Degano, P., Zavattaro, G. (eds.) SFM 2008. LNCS, vol. 5016,
pp. 215264. Springer, Heidelberg (2008)

Hybrid Petri Nets for Modelling the Eukaryotic Cell Cycle

141

9. Heiner, M., Lehrack, S., Gilbert, D., Marwan, W.: Extended stochastic petri nets
for model-based design of wetlab experiments. In: Priami, C., Back, R.-J., Petre,
I. (eds.) Transactions on Computational Systems Biology XI. LNCS, vol. 5750, pp.
138163. Springer, Heidelberg (2009)
10. Herajy, M., Heiner, M.: Hybrid representation and simulation of sti biochemical
networks. Nonlinear Analysis: Hybrid Systems 6(4), 942959 (2012)
11. Herajy, M.: Computational Steering of Multi-Scale Biochemical Reaction Networks. Ph.D. thesis, Brandenburg University of Technology Cottbus - Computer
Science Institute (2013)
12. Herajy, M., Heiner, M.: Hybrid representation and simulation of sti biochemical
networks through generalised hybrid Petri nets. Tech. Rep. 02/2011, Brandenburg
University of Technology Cottbus, Dept. of CS (2011)
13. Kar, S., Baumann, W.T., Paul, M.R., Tyson, J.J.: Exploring the roles of noise in
the eukaryotic cell cycle. Proceedings of the National Academy of Sciences of the
United States of America 106(16), 64716476 (2009)
14. Kiehl, T., Mattheyses, R., Simmons, M.: Hybrid simulation of cellular behavior.
Bioinformatics 20, 316322 (2004)
15. Liu, Z., Pu, Y., Li, F., Shaer, C., Hoops, S., Tyson, J., Cao, Y.: Hybrid modeling
and simulation of stochastic eects on progression through the eukaryotic cell cycle.
J. Chem. Phys. 136(34105) (2012)
16. Marwan, W., Rohr, C., Heiner, M.: Petri nets in Snoopy: A unifying framework
for the graphical display, computational modelling, and simulation of bacterial
regulatory networks. Methods in Molecular Biology, ch. 21, vol. 804, pp. 409437.
Humana Press (2012)
17. Matsui, M., Fujita, S., Suzuki, S., Matsuno, H., Miyano, S.: Simulated cell division
processes of the xenopus cell cycle pathway by genomic object net. Journal of
Integrative Bioinformatics, 0001 (2004)
18. Matsuno, H., Tanaka, Y., Aoshima, H., Doi, A., Matsui, M., Miyano, S.: Biopathways representation and simulation on hybrid functional Petri net. Silico Biology 3(3) (2003)
19. Mura, I., Csik
asz-Nagy, A.: Stochastic Petri net extension of a yeast cell cycle
model. Journal of Theoretical Biology 254(4), 850860 (2008)
20. Sabouri-Ghomi, M., Ciliberto, A., Kar, S., Novak, B., Tyson, J.J.: Antagonism and
bistability in protein interaction networks. Journal of Theoretical Biology 250(1),
209218 (2008)
21. Singhania, R., Sramkoski, R.M., Jacobberger, J.W., Tyson, J.J.: A hybrid model
of mammalian cell cycle regulation. PLoS Comput. Biol. 7(2), e1001077 (2011)
22. Steuer, R.: Eects of stochasticity in models of the cell cycle: from quantized cycle
times to noise-induced oscillations. Journal of Theoretical Biology 228(3), 293301
(2004)
Nov
23. Sveiczer, A.,
ak, B., Mitchison, J.: The size control of ssion yeast revisited. J.
Cell Sci. 109, 29472957 (1996)
24. Tyson, J., Novak, B.: Regulation of the eukaryotic cell cycle: Molecular antagonism,
hysteresis, and irreversible transitions. Journal of Theoretical Biology 210(2), 249
263 (2001)
25. Tyson, J., Novak, B.: A Systems Biology View of the Cell Cycle Control Mechanisms. Elsevier, San Diego (2011)
26. Valk, R.: Self-modifying nets, a natural extension of Petri nets. In: Ausiello, G.,
B
ohm, C. (eds.) Proceedings of the Fifth Colloquium on Automata, Languages
and Programming, vol. 62, pp. 464476. Springer, Heidelburg (1978)
27. Wikipedia: Wikipedia website (2012), http://www.wikipedia.org/ (accessed:
September 20, 2012)