TWMS J. App. Eng. Math. V.3, N.2, 2013, pp. 182-197.

THE METHOD OF LINES FOR THE NUMERICAL SOLUTION OF A

MATHEMATICAL MODEL IN THE INITIATION OF ANGIOGENESIS

S. PAMUK1 , I. ATAC2 §

Abstract. In this paper we present the method of lines to obtain the numerical solution
of a mathematical model for the roles of endothelial, pericyte and macrophage cells in the
initiation of tumor angiogenesis. This method is an approach to the numerical solution
of partial differential equations that involve a time variable t and a space variable x. We
provide computer programs written in Matlab. Also, the stability analysis of the solutions
is given, and the figures for endothelial, pericyte and macrophage cell movements in the
capillary are presented for large times.

Keywords: Method of lines, stability, angiogenesis, capillary

AMS Subject Classification: 65M20, 34D20

1. Introduction
Angiogenesis is the main feature of the formation of new blood vessels. The abluminal
surface of the capillaries is covered by a collageneous network intermingled with laminin
which is called the basal lamina (BL). The BL is mainly formed by the Endothelial Cells
(EC). In the neighborhood of BL, there are other cell types such as Pericyte Cells (PC)
and Macrophage Cells (MC) [4].
In this paper we consider the following initial boundary-value problem originally pre-
sented in [4]: ( ( ))
∂η ∂ ∂ η
= D1 η ln , 0 < x < 1, 0 < t ≤ T, (1)
∂t ∂x ∂x τ1
( ( ))
∂σ ∂ ∂ σ
= D2 σ ln , 0 < x < 1, 0 < t ≤ T, (2)
∂t ∂x ∂x τ
( ( 2 ))
∂m ∂ ∂ m
= D3 m ln , 0 < x < 1, 0 < t ≤ T, (3)
∂t ∂x ∂x τ3
where η(x, t), σ(x, t) and m(x, t) are the concentrations of EC, PC and MC, respectively,
and τ1 , τ2 and τ3 are the transition probability functions of them. Also, D1 , D2 , D3 are
EC, PC and MC diffusion constants. We impose the initial conditions,
η(x, 0) = 1, σ(x, 0) = 1, m(x, 0) = 1, 0 < x < 1, (4)
and the boundary conditions,

1
Department of Mathematics, University of Kocaeli, Umuttepe Campus, 41380, Kocaeli - TURKEY,
e-mail: [email protected]
2
Department of Mathematics, University of Kocaeli, Umuttepe Campus, 41380, Kocaeli - TURKEY,
e-mail: [email protected]
§ Manuscript received April 04, 2013.
TWMS Journal of Applied and Engineering Mathematics Vol.3, No.1; ⃝
c Işık University, Department
of Mathematics, 2013; all rights reserved.
182
 S. PAMUK, I. ATAC: THE METHOD OF LINES... 183
( ) ( ) ( )
∂ η ∂ σ ∂ m
D1 η ln = D2 σ ln = D3 m ln = 0, (5)
∂x τ1 ∂x τ2 ∂x τ3
( ) x=0 ( ) x=0 ( ) x=0
∂ η ∂ σ ∂ m
D1 η ln = D2 σ ln = D3 m ln = 0. (6)
∂x τ1 x=1 ∂x τ2 x=1 ∂x τ3 x=1
As in [6,7], for numerical purposes we take the transition probability functions τ1 , τ2
and τ3 as follows:
( ) ( )
α1 + Axn (1 − x)n γ1 β1 + 1 − Bxn (1 − x)n γ2
τ1 = f (x) = , (7)
α2 + Axn (1 − x)n β2 + 1 − Bxn (1 − x)n
( )
α3 + 1 − Bxn (1 − x)n γ3
τ2 = g(x) = , (8)
α4 + 1 − Bxn (1 − x)n
( )
β3 + Axn (1 − x)n γ4
τ3 = r(x) = . (9)
β4 + Axn (1 − x)n
Here, n, A, B, αi , βi and γi (i = 1 : 4) are some positive constants which we take here
to be the same as in [4]. The Authors of Ref.[4] have chosen the transition probability
functions f (x), g(x) and r(x) more complicated for biological reasons. But for simplicity,
we take them as a function of x only.

2. Method
The method of lines (MOL) is a general way of viewing a partial differential equation
(PDE) as a system of ordinary differential equations (ODE) [1-3,5,6,9-11]. The partial
derivatives with respect to the space variables are discretized to obtain a system of ODEs
in the time variable and then a suitable initial-value problem solver is used to solve this
ODE system. This method has the broad applicability to physical and chemical systems
modeled by PDEs [9]. Also, this method provides very accurate numerical solution for
linear or nonlinear PDE’ s in comparison with other existing methods [6]. As the number
of lines increases, the accuracy of the MOL representation of the original system increases
[5]. We now apply this method to our problem given by Eqs.(1)-(6).
We proceed uniform grid,
W = {(xi , tj ) : xi = (i − 1)h, tj = (j − 1)k, i = 1 : M, j = 1 : N, h = 1/(M − 1), k =
1/(N − 1)} and obtain a difference scheme for the problem in Eqs.(1)-(6). Eqs.(1)-(3) can
be written as follows:
ηt = D1 (ηxx − (ηF )x ) , (10)

σt = D2 (σxx − (σG)x ) , (11)

mt = D1 (mxx − (mH)x ) , (12)

f ′ (x) g ′ (x) r′ (x)

where F = ,G= and H = .
f (x) g(x) r(x)
We suppose the solutions of Eqs.(10)-(12) η(x, t), σ(x, t) and m(x, t) can be approxi-
mated by ηi , σi and mi , respectively, and use the standart difference equations for the first
and the second derivatives of these variables. Therefore, by discretizing the right handside
of Eqs.(10)-(12)
( we obtain the following system of ODE’s: )
dηi ηi+1 − 2ηi + ηi−1 ηi+1 Fi+1 − ηi−1 Fi−1
= D1 − , (13)
dt h2 2h
184 TWMS J. APP. ENG. MATH. V.3, N.2, 2013
( )
dσi σi+1 − 2σi + σi−1 σi+1 Gi+1 − σi−1 Gi−1
= D2 − , (14)
dt h2 2h
( )
dmi mi+1 − 2mi + mi−1 mi+1 Hi+1 − mi−1 Hi−1
= D3 − . (15)
dt h2 2h
The initial conditions in Eq.(4) and boundary conditions in Eqs.(5)-(6) now become:
ηi = 1, σi = 1, mi = 1, 1 < i < M , t = 0, (16)
( )
∂η1
D1 − η1 F1 = 0, for t > 0, (17)
∂x
( )
∂ηM
D1 − ηM FM = 0, for t > 0, (18)
∂x
( )
∂σ1
D2 − σ1 G1 = 0, for t > 0, (19)
∂x
( )
∂σM
D2 − σM GM = 0, for t > 0, (20)
∂x
( )
∂m1
D3 − m1 H1 = 0, for t > 0, (21)
∂x
( )
∂mM
D3 − mM HM = 0, for t > 0. (22)
∂x
Using the central difference for the boundary conditions in Eqs.(17)-(22) and taking
into account Eqs.(13)-(15), we obtain the following system of ODE’s:

( )
dη1 2η2 − 2(1 + hF1 )η1 η2 F2 − (η2 − 2hη1 F1 )F0
= D1 − , (23)
dt h2 2h
( )
dηi ηi+1 − 2ηi + ηi−1 ηi+1 Fi+1 − ηi−1 Fi−1
= D1 − , 1 < i < M, (24)
dt h2 2h
( )
dηM 2ηM −1 − 2(1 − hFM )ηM (ηM −1 + 2hηM FM )FM +1 − ηM −1 FM −1
= D1 2
− , (25)
dt h 2h
( )
dσ1 2σ2 − 2(1 + hG1 )σ1 σ2 G2 − (σ2 − 2hσ1 G1 )G0
= D2 − , (26)
dt h2 2h
( )
dσi σi+1 − 2σi + σi−1 σi+1 Gi+1 − σi−1 Gi−1
= D2 − , 1 < i < M, (27)
dt ( h2 2h )
dσM 2σM −1 − 2(1 − hGM )σM (σM −1 + 2hσM GM )GM +1 − σM −1 GM −1
= D2 2
− ,(28)
dt h 2h
( )
dm1 2m2 − 2(1 + hH1 )m1 m2 H2 − (m2 − 2hm1 H1 )H0
= D3 − , (29)
dt h2 2h
( )
dmi mi+1 − 2mi + mi−1 mi+1 Hi+1 − mi−1 Hi−1
= D3 − , 1 < i < M, (30)
dt h2 2h
( )
dmM 2mM −1 − 2(1 − hHM )mM (mM −1 + 2hmM HM )HM +1 − mM −1 HM −1
= D3 − .
dt h2 2h
(31)
 S. PAMUK, I. ATAC: THE METHOD OF LINES... 185

3. Stability of the solutions

We let
x = [η1 · · · ηi · · · ηM σ1 · · · σi · · · σM m1 · · · mi · · · mM ]T (32)
and  ( ) 
D1 2η2 −2(1+hF h2
1 )η1
− η2 F2 −(η2 −2hη 1 F1 )F0

 2h

 .. 
 ( . ) 
 
 ηi+1 −2ηi +ηi−1
− ηi+1 Fi+1 −ηi−1 Fi−1 
 D 1 h2 2h 
 
 .. 
 ( . ) 
 
 D1 2ηM −1 −2(1−hF M )ηM
−
(ηM −1 +2hηM FM )FM +1 −ηM −1 FM −1 
 (h
2 2h ) 
 
 D2 2σ2 −2(1+hG1 )σ1
− σ2 G2 −(σ2 −2hσ1 G1 )G0 
 h2 2h 
 
 .
.. 
 ( ) 
 
f(x) = 
 D2 σ i+1 −2σ
h2
i +σ i−1
− σ i+1 G i+1 −σ
2h
i−1 G i−1 ,
 (33)
 
 .. 
 ( . ) 
 
 D2 2σM −1 −2(1−hG M )σM (σ
− M −1
+2hσM GM )GM +1 −σM −1 GM −1 
 ( h2 2h ) 
 2m2 −2(1+hH1 )m1 m2 H2 −(m2 −2hm1 H1 )H0 
 D3 − 
 h2 2h 
 .. 
 
 ( . ) 
 mi+1 −2mi +mi−1 mi+1 Hi+1 −mi−1 Hi−1 
 D3 − 
 h2 2h 
 .. 
 . 
 ( ) 
2mM −1 −2(1−hHM )mM (m +2hmM HM )HM +1 −mM −1 HM −1
D3 h2
− M −1 2h
where 2 ≤ i ≤ M − 1. Then, by Eqs.(23)-(31) one has the following autonomous system
of ordinary differential equations:

ẋ = f(x), t ≥ 0. (34)

We now let

η(x, t) = 1 + η ∗ (x, t), σ(x, t) = 1 + σ ∗ (x, t), m(x, t) = 1 + m∗ (x, t), (35)
′ ′ ′
and assume that F (x) ≈ 0, G (x) ≈ 0, H (x) ≈ 0 for small x. These assumptions are
reasonable (see Figures (1)-(3)). Therefore, Eqs.(10)-(15) and Eqs.(17)-(31) stay the same
in the variables with the asterisks except the initial conditions in Eq.(16) now become
ηi∗ = 0, σi∗ = 0, m∗i = 0, 1 < i < M , t = 0. (36)

For simplicity we drop all of the asterisks. Therefore, we now solve Eqs.(23)-(31) with
the initial conditions in Eq.(36).
It is now clear to see that the only steady-state solution to Eqs.(23)-(31) is the vector
c = (0 · · · 0 0 · · · 0 0 · · · 0)T of length 3M . Let A, B and C be the matrices of first
order partial derivatives of f(x) with respect to the variables ηi , σi and mi (1 ≤ i ≤ M ),
respectively, evaluated at c. Therefore, the matrices A, B and C become
186 TWMS J. APP. ENG. MATH. V.3, N.2, 2013

 2(1+hF1 ) F2 F0

−D1 ( h2
+ F1 F0 ) D1 ( h22 − + ) 0 ··· 0
 2h 2h

 . 
 D1 ( h12 + F1
) − 2D 1
D1 ( h12 − F3
) .. 
 2h h2 2h 
 
A=

..
.
..
.
..
.


 0 0 
 . 
 .. FM −2 
 D1 ( h12 + 2h
) − 2D
h2
1
D1 ( h12 − F2h
M
) 
2hFM −2
0 ··· 0 0 D 1 ( h2 − FM FM +1 )

 2(1+hG1 ) G2 G0

−D2 ( h2
+ G1 G0 ) D2 ( h22 − + ) 0 ··· 0
 2h 2h

 .. 
 D2 ( h12 + G1
) − 2D 2
D2 ( h12 − G3
) . 
 2h h2 2h 
 
B=

..
.
..
.
..
.


 0 0 
 . 
 .. GM −2 
 D2 ( h12 + 2h
) − 2D
h2
2
D2 ( h12 − G2h
M
) 
2hGM −2
0 ··· 0 0 D2 ( h2
− GM GM +1 )

 2(1+hH1 ) H2 H0

−D3 ( h2
+ H1 H0 ) D3 ( h22 − + ) 0 ··· 0
 2h 2h

 . 
 D3 ( h12 + H1
) − 2D 3
D3 ( h12 − H3
) .. 
 2h h2 2h 
 
C=

..
.
..
.
..
.


 0 0 
 .. 
 HM −2 
 . D3 ( h12 + 2h
) − 2D
h2
3
D3 ( h12 − H2h
M
) 
−2
0 ··· 0 0 D3 ( 2hHhM
2 − H M HM +1 )

We let  
A O
S= B 
O C
where O is the zero matrix. It is well known that the determinant of the matrix S is
the product of the determinants of the matrices A, B and C. Therefore, the eigenvalues
of S consists of the set of eigenvalues of A, B and C. We know that every solution of
the system of equations given by Eqs.(23)-(31) is stable if all the eigenvalues of S have
negative real part and unstable if at least one eigenvalue of S has positive real part [12].
In the following section we solve the system in Eqs.(23)-(31) with the initial conditions
in Eq.(36) by using an ordinary differential equation solver which is built up in Matlab,
and discuss the stability (or unstability) of the solutions of the system by determining the
signs of the real parts of the eigenvalues of the matrix S for some values of M .

4. Numerical experiments
First, we solve the system in Eqs.(23)-(31) by the method of lines. For numerical
purposes we take the following constants and parameters from [4,6,7]. D1 = D2 = D3 =
0.00025, α1 = 0.1, α2 = 2, α3 = 0.15, α4 = 1, β1 = 10, β2 = 0.1, β3 = 0.5, β4 = 1, γ1 =
1, γ2 = 1, γ3 = 1, γ4 = 1, n = 16, A = 28 × 107 , B = 0.22 × 109 . We first code the input data
as global variables and we code the system Eqs.(23)-(31) as a m-file called ”fparabolic”.
Then, we use the form [t,y]=ode45(’fparabolic’, tspan, init) to obtain the solution. Here,
”ode45” is an ODE solver which is built up in Matlab, ”fparabolic” is a function m file we
have just described above, ”tspan” is the desired time vector which we want the solution
for, and ”init” is the initial condition vector. Also, ”t” is the same as tspan and ”y” is
the solution vector of the system. The first M elements of y give EC concentration, the
second M elements of y give PC concentration and the third M elements of y give MC
 S. PAMUK, I. ATAC: THE METHOD OF LINES... 187

concentration (see Section 5.1 for Matlab codes). These concentrations are shown in Figs.
4-7, respectively. Furthermore, the importance of these concentrations in the initiation of
angiogenesis is explained in Conclusions.
Second, we test the signs of the real parts of the eigenvalues of the matrix S for the
system in Eqs.(23)-(31) (see Section 5.2 for the construction of the matrix S in Matlab).
We do this by assuming F ′ (x) ≈ 0, G′ (x) ≈ 0, H ′ (x) ≈ 0 for small x, and taking
M = 21, 23, 25, 27, 29, 31, 41, and 45, respectively. When M = 21, 23, 25, 27, 29, 31 we
observe that the real parts of the eigenvalues of S are negative (see Table 1), and thus
all of the solutions of our system are stable for large time levels [12]. Figs. 4-6 are
some snapshots taken at t = 10, 110, 210, 310, 410, 510, 610 and t = 710. This is a strong
evidence that all of the solutions of our system are stable for all time levels. But, when
M = 41 and M = 45 we see that the real parts of two of the eigenvalues of S are positive
(see Table 1), which means that every solution of our system is unstable. As it is clear
from the Fig.7 that the stability changes to unstability around t = 589. As users may
remember the ODE solver ”ode45” which is built up in Matlab chooses its time step ”k”
automatically. One of the reason we get unstability for M = 45 as time increases may be
the violation of the CFL condition
Di k 1
2
≤ , (i = 1, 2, 3),
h 2
where Di are the diffusion constants and h = 1/(M − 1). As M increases one has to
take k very small to meet the CFL condition since Di are fixed constants. Therefore, the
CFL condition is a severe restriction on time step.
The other reason for the unstability for M = 45 case could be the rounding errors in
our computations. Although rounding errors can excite an unstability, more often it is
discretisation errors that do so.
Fig.7 would have looked the same even if the computations had been carried out by
using stiff ODE solvers built up in Matlab such as ode15s, ode23s.
Of course, a fully implicit or the Crank-Nicolson method can be used to address the
unstability for most of the problems in PDEs. Although there is no restriction on time
step for both methods, they require more work at each time step. Especially the Crank-
Nicolson method is unconditionally stable.

0.25 2

0.2
1
0.15

0.1
0
0.05
F’

0 −1
F

−0.05
−2
−0.1

−0.15
−3
−0.2

−0.25 −4
0 0.5 1 0 0.5 1
x x

f ′ (x)
Figure 1. F (x) = f (x) and F ′ (x).
188 TWMS J. APP. ENG. MATH. V.3, N.2, 2013

0.2 4

0.15
3

0.1

2
0.05

G’
G

0 1

−0.05
0

−0.1

−1
−0.15

−0.2 −2
0 0.5 1 0 0.5 1
x x

g ′ (x)
Figure 2. G(x) = g(x) and G′ (x).

0.4 3

0.3 2

1
0.2

0
0.1
−1
H’
H

0
−2
−0.1
−3

−0.2
−4

−0.3 −5

−0.4 −6
0 0.5 1 0 0.5 1
x x

r ′ (x)
Figure 3. H(x) = r(x) and H ′ (x).

1.5

1.4

1.3

1.2
η

1.1

0.9

0.8
0 0.2 0.4 0.6 0.8 1
x

Figure 4. EC Concentration η(x, t) with M = 31.

 S. PAMUK, I. ATAC: THE METHOD OF LINES... 189

1.005

0.995
σ

0.99

0.985

0.98
0 0.2 0.4 0.6 0.8 1
x

Figure 5. PC Concentration σ(x, t) with M = 31.

1.05

1.04

1.03

1.02
m

1.01

0.99

0.98
0 0.2 0.4 0.6 0.8 1
x

Figure 6. MC Concentration m(x, t) with M = 31.

t=584 t=589
1.005 1.005
1 1
0.995 0.995
σ

0.99 0.99
0.985 0.985
0.98 0.98
0 0.5 1 0 0.5 1
x x
t=673 t=698
1.005 1.005
1 1
0.995 0.995
σ

0.99 0.99
0.985 0.985
0.98 0.98
0 0.5 1 0 0.5 1
x x

Figure 7. PC Concentration σ(x, t) with M = 45.

190 TWMS J. APP. ENG. MATH. V.3, N.2, 2013

M=21 M=23 M=25 M=27 M=29 M=31 M=41 M=45

-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.2000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.0000 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.0030 -0.2420 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.0087 -0.4840 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.0231 -0.4810 -0.2880 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.0379 -0.4753 -0.0000 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.0827 -0.4608 -0.0030 -0.3380 -0.3920 -0.4500 -0.8000 -0.9680
-0.0587 -0.4459 -0.0087 -0.6760 -0.3920 -0.4500 -0.8000 -0.9680
-0.1092 -0.4247 -0.0233 -0.6730 -0.3920 -0.4500 -0.8000 -0.9680
-0.1383 -0.4002 -0.0383 -0.6673 -0.0000 -0.4500 -0.8000 -0.9680
-0.1687 -0.3728 -0.0597 -0.6527 -0.0030 -0.4500 -0.8000 -0.9680
-0.2313 -0.3425 -0.1127 -0.0000 -0.0234 -0.9000 -0.8000 -0.9680
-0.4000 -0.3101 -0.0846 -0.0030 -0.0385 -0.8970 -0.8000 -0.9680
-0.3970 -0.2764 -0.1441 -0.0087 -0.0858 -0.8766 -0.8000 -0.9680
-0.3913 -0.0000 -0.1779 -0.6376 -0.1149 -0.8614 -0.8000 -0.9680
-0.3769 -0.0030 -0.2135 -0.0600 -0.1477 -0.0000 -0.8000 -0.9680
-0.3621 -0.0087 -0.2504 -0.0384 -0.7840 -0.0030 -0.8000 -0.9680
-0.2617 -0.0232 -0.3256 -0.0853 -0.7810 -0.0234 -0.8000 -0.9680
-0.2908 -0.0381 -0.3625 -0.6160 -0.7606 -0.0386 -0.8000 -0.9680
-0.3173 -0.0593 -0.5760 -0.1139 -0.7753 -0.8395 -0.8000 -0.9680
-0.3413 -0.0838 -0.5730 -0.3787 -0.7455 -0.4970 -0.8000 -0.9680
-0.0586 -0.1112 -0.5673 -0.2973 -0.2220 -0.4030 -1.6000 -0.9680
-0.3414 -0.1415 -0.5527 -0.5907 -0.2626 -0.6330 -1.5970 -0.9680
-0.2000 -0.1739 -0.5377 -0.1461 -0.5214 -0.5890 -1.5913 -0.9680
-0.0000 -0.2076 -0.3981 -0.2572 -0.5620 -0.2670 -1.5765 -0.9680
-0.0025 -0.4131 -0.5163 -0.4188 -0.6982 -0.3110 -1.5611 -1.9360
 S. PAMUK, I. ATAC: THE METHOD OF LINES... 191

M=21 M=23 M=25 M=27 M=29 M=31 M=41 M=45

-0.0097 -0.0709 -0.4319 -0.5299 -0.3481 -0.0605 -1.5389 -1.9330
-0.0219 -0.2420 -0.4633 -0.4578 -0.4359 -0.6749 -1.5125 -1.9273
-0.0381 -0.0000 -0.4914 -0.5621 -0.1835 -0.2251 -1.4820 -1.9125
-0.0586 -0.0025 -0.0844 -0.1810 -0.0602 -0.7510 -1.4471 -1.8971
-0.0824 -0.0097 -0.2880 -0.2182 -0.6005 -0.7843 -1.4082 -1.8748
-0.1092 -0.0220 -0.4916 -0.4950 -0.0087 -0.1490 -1.3194 -1.8482
-0.1382 -0.0384 -0.5760 -0.0233 -0.6363 -0.1157 -1.1061 -1.8175
-0.1687 -0.0591 -0.5735 -0.5770 -0.7238 -0.8138 -1.1631 -1.7822
-0.2000 -0.0835 -0.5663 -0.3380 -0.3048 -0.0862 -1.0472 -1.7428
-0.2313 -0.1112 -0.5540 -0.0990 -0.4792 -0.3565 -0.9251 -1.6994
-0.2618 -0.1415 -0.5375 -0.6760 -0.6691 -0.5435 -0.8628 -1.6018
-0.2908 -0.1738 -0.5165 -0.6735 -0.6692 -0.1856 0.0000 -1.5480
-0.3176 -0.2076 -0.4916 -0.6663 -0.3920 -0.7144 -0.0030 -1.4912
-0.4000 -0.2420 -0.4633 -0.6539 -0.1148 -0.0087 -0.4939 -1.4318
-0.3975 -0.2764 -0.4320 -0.6373 -0.7840 -0.8913 -0.5528 -1.3701
-0.3903 -0.3102 -0.3982 -0.6162 -0.7815 -0.7682 -0.4369 -1.3062
-0.3781 -0.3425 -0.3625 -0.5910 -0.7743 -0.1318 -0.0087 -1.2407
-0.3619 -0.3728 -0.3256 -0.5621 -0.7619 -0.4500 -0.0611 -1.1737
-0.3414 -0.4840 -0.2880 -0.4579 -0.7452 -0.9000 -0.0389 -1.1057
-0.4815 -0.2504 -0.4189 -0.7239 -0.8975 -1.2701 -1.0370
-0.4743 -0.2135 -0.4951 -0.6985 -0.8903 -0.0875 -0.8990
-0.4620 -0.1778 -0.3787 -0.0000 -0.8779 -0.6749 -0.8303
-0.4456 -0.1440 -0.2973 -0.6692 -0.8612 -0.3821 -0.7623
-0.4005 -0.1127 -0.2571 -0.4359 -0.8397 -1.3656 -0.6953
-0.4249 -0.0000 -0.2181 -0.6364 -0.8141 -0.1918 -0.6298
-0.0025 -0.1809 -0.5621 -0.7511 -0.1180 -0.5659
-0.0097 -0.1139 -0.5215 -0.7145 -0.1529 -0.5042
-0.0220 -0.0000 -0.4792 -0.7844 -1.2179 0.0000
-0.0385 -0.0025 -0.3481 -0.0000 -0.0235 -0.0030
-0.0595 -0.0850 -0.3048 -0.6330 -0.2806 -0.0087
-0.0844 -0.0097 -0.2219 -0.4970 -0.7372 -0.0235
-0.0387 -0.2625 -0.5436 -0.9867 -0.0389
-0.0598 -0.6006 -0.4030 -0.6133 -0.0612
-0.5300 -0.1148 -0.3564 -0.3299 -0.0878
-0.0221 -0.0097 -0.6750 -0.2344 -0.1185
-0.1460 -0.0221 -0.2670 -1.3657 -0.1538
-0.3380 -0.0601 -0.5891 -0.8000 -0.1932
-0.1476 -0.1855 -0.2343 -0.4448
-0.0855 -0.0221 -1.6000 -0.2366
-0.1834 -0.0097 -1.5975 -0.3880
-0.0388 -0.3109 -1.5903 -0.3342
-0.0025 -0.0859 -1.5778 -1.6524
-0.3920 -0.2250 -1.5609 -0.2836
-0.1489 -1.5391 -1.6525
-0.0603 -1.5128 -0.9680
-0.0025 -1.4821 -0.2835
192 TWMS J. APP. ENG. MATH. V.3, N.2, 2013

M=21 M=23 M=25 M=27 M=29 M=31 M=41 M=45

-0.1156 -1.4472 0.0000
-0.0388 -1.3657 -0.0025
-0.4500 -1.3196 -0.0097
-1.2702 -0.0222
-1.1632 -0.0391
-0.8628 -0.0610
-0.9251 -0.0875
-1.1061 -0.1184
-0.7372 -0.1537
-0.9868 -0.1931
-0.6749 -0.2364
0.0000 -0.2835
-0.6132 -0.3341
-0.4939 -0.3879
-0.4368 -0.4447
-0.0025 -0.5041
-0.3298 -0.5659
-0.1179 -0.6297
-0.0391 -0.6953
-0.0222 -0.7622
-0.1528 -0.8302
-0.2804 -0.8989
-0.2343 -0.9680
-0.0872 -1.0371
-1.0472 -1.1058
-1.2180 -1.1738
-0.5528 -1.2407
-0.0609 -1.3063
-0.3820 -1.3701
-1.4083 -1.4319
-0.0097 -1.4913
-0.1917 -1.5481
-0.8000 -1.6019
-1.9360
-1.9335
-1.9263
-1.9138
-1.8969
-1.8750
-1.8485
-1.8176
-1.7823
-1.7429
-1.6996
-1.6525
Table 1: Real Parts of the Eigenvalues of the Matrix S for some Values of M

5. Conclusions
In this paper we have presented the MOL for the numerical solution of a mathematical
model for the roles of EC, PC and MC in the initiation of tumor angiogenesis. This
method does not require large computer memory, and avoids linearization and physically
 S. PAMUK, I. ATAC: THE METHOD OF LINES... 193

unrealistic assumptions. As the number of lines increases, the accuracy of the MOL
representation increases. This method is also applicable to non-linear systems of PDEs.
In Figs. 4-6, we have plotted the EC, PC and MC concentrations for large times. Notice
that pericyte aggregation in the opening in the capillary wall is low where the EC and
MC concentrations are high and conversely. In particular, the maximum of PC density
is higher closer to the wall of the forming capillary than is the maximum of EC density
[4]. These are consistent with the observations of [8] that the PC closely regulate the
development of the forming channel.
We have also showed that when M = 21, 23, 25, 27, 29, 31 all of the solutions of our
system obtained from the discretisation of the space variable x are stable. This a strong
evidence that all of the solutions are stable for all time levels as it is clear from the Figs. 4-
6. To show the stability of the solutions, we have obtained (in Table 1) the negativity of the
real parts of the eigenvalues of the matrix (namely S) consisting of the first order partial
derivatives of the discretised vectoral function with respect to the unknown variables,
evaluated at a constant steady state solution.
When we increase M we have faced that the real parts of two of the eigenvalues of the
matrix S are positive, which means that every solution of our system is unstable. For
example, when M = 45 the stability changes to unstability around t = 589 as it is clear
from the Fig. 7, which may be caused by the violation of the CFL condition.
We have also presented our Matlab codes in the Appendix A and B below, which
may enlighten young researchers to learn one of the numerical methods to solve PDEs or
systems of PDEs.

6. Appendix A
Matlab codes for the method of lines
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function z=fparabolic(t,w)
global h M
alpha1=0.1; alpha2=2; alpha3=0.15; alpha4=1; beta1=10; beta2=0.1; beta3=0.5; beta4=1;
gama1=1;
gama2=1; gama3=1; gama4=1; D1=0.00025; D2=0.00025; D3=0.00025; n=16;
A=28*10ˆ7; B=0.22*10ˆ9;
eta=w(1:M); sigma=w(M+1:2*M); m=w(2*M+1:3*M);
h=1./(M-1);
%%%%%%%%%%%%%% F,G,H functions%%%%%%%%%%%%%%%
for i=1:M
x(i)=(i-1)*h; % the vector x
end
F0=gama1*(alpha2-alpha1)*A*n*(-h-hˆ2)ˆ(n-1)*(1+2*h)/...
((alpha1+A*(-h-hˆ2)ˆn)*(alpha2+A*(-h-hˆ2)ˆn))+...
gama2*(beta1-beta2)*B*n*(-h-hˆ2)ˆ(n-1)*(1+2*h)/...
((beta1+1-B*(-h-hˆ2)ˆn)*(beta2+1-B*(-h-hˆ2)ˆn)); % F(x) at x=0
G0=gama3*(alpha3-alpha4)*B*n*(-h-hˆ2)ˆ(n-1)*(1+2*h)/...
((alpha3+1-B*(-h-hˆ2)ˆn)*(alpha4+1-B*(-h-hˆ2)ˆn)); % G(x) at x=0
H0=gama4*(beta4-beta3)*A*n*(-h-hˆ2)ˆ(n-1)*(1+2*h)/...
((beta3+A*(-h-hˆ2)ˆn)*(beta4+A*(-h-hˆ2)ˆn)); % H(x) at x=0
for i=1:M
F(i)=gama1*(alpha2-alpha1)*A*n*(x(i)-x(i)ˆ2)ˆ(n-1)*(1-2*x(i))/...
((alpha1+A*(x(i)-x(i)ˆ2)ˆn)*(alpha2+A*(x(i)-x(i)ˆ2)ˆn))+...
194 TWMS J. APP. ENG. MATH. V.3, N.2, 2013

gama2*(beta1-beta2)*B*n*(x(i)-x(i)ˆ2)ˆ(n-1)*(1-2*x(i))/...
((beta1+1-B*(x(i)-x(i)ˆ2)ˆn)*(beta2+1-B*(x(i)-x(i)ˆ2)ˆn)); % F(x) in the interval
0<x<1
G(i)=gama3*(alpha3-alpha4)*A*n*(x(i)-x(i)ˆ2)ˆ(n-1)*(1-2*x(i))/...
((alpha3+1-A*(x(i)-x(i)ˆ2)ˆn)*(alpha4+1-A*(x(i)-x(i)ˆ2)ˆn)); % G(x) in the inter-
val 0<x<1
H(i)=gama4*(beta4-beta3)*B*n*(x(i)-x(i)ˆ2)ˆ(n-1)*(1-2*x(i))/...
((beta3+B*(x(i)-x(i)ˆ2)ˆn)*(beta4+B*(x(i)-x(i)ˆ2)ˆn)); % H(x) in the interval 0<x<1
end
F(M+1)=gama1*(alpha2-alpha1)*A*n*(M*h-(M*h)ˆ2)ˆ(n-1)*(1-2*M*h)/...
((alpha1+A*(M*h-(M*h)ˆ2)ˆn)*(alpha2+A*(M*h-(M*h)ˆ2)ˆn))+...
gama2*(beta1-beta2)*B*n*(M*h-(M*h)ˆ2)ˆ(n-1)*(1-2*M*h)/...
((beta1+1-B*(M*h-(M*h)ˆ2)ˆn)*(beta2+1-B*(M*h-(M*h)ˆ2)ˆn)); % F(x) at x=1
G(M+1)=gama3*(alpha3-alpha4)*A*n*(M*h-(M*h)ˆ2)ˆ(n-1)*(1-2*M*h)/...
((alpha3+1-A*(M*h-(M*h)ˆ2)ˆn)*(alpha4+1-A*(M*h-(M*h)ˆ2)ˆn)); % G(x) at x=1
H(M+1)=gama4*(beta4-beta3)*B*n*(M*h-(M*h)ˆ2)ˆ(n-1)*(1-2*M*h)/...
((beta3+B*(M*h-(M*h)ˆ2)ˆn)*(beta4+B*(M*h-(M*h)ˆ2)ˆn)); % H(x) at x=1
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
z(1)=D1*((2*eta(2)-2*eta(1)-2*h*eta(1)*F(1))/hˆ2-...
(eta(2)*F(2)-(eta(2)-2*h*eta(1)*F(1))*F0)/(2*h)); % gives Eq. (23)
z(M+1)=D2*((2*sigma(2)-2*sigma(1)-2*h*sigma(1)*G(1))/hˆ2-...
(sigma(2)*G(2)-(sigma(2)-2*h*sigma(1)*G(1))*G0)/(2*h)); % gives Eq. (26)
z(2*M+1)=D3*((2*m(2)-2*m(1)-2*h*m(1)*H(1))/hˆ2-...
(m(2)*H(2)-(m(2)-2*h*m(1)*H(1))*H0)/(2*h)); % gives Eq. (29)
for i=2:M-1;
z(i)=D1*((eta(i+1)-2*eta(i)+eta(i-1))/hˆ2-...
(eta(i+1)*F(i+1)-eta(i-1)*F(i-1))/(2*h)); % gives Eq. (24)
z(M+i)=D2*((sigma(i+1)-2*sigma(i)+sigma(i-1))/hˆ2-...
(sigma(i+1)*G(i+1)-sigma(i-1)*G(i-1))/(2*h)); % gives Eq. (27)
z(2*M+i)=D3*((m(i+1)-2*m(i)+m(i-1))/hˆ2-...
(m(i+1)*H(i+1)-m(i-1)*H(i-1))/(2*h)); % gives Eq. (30)
end
z(M)=D1*((2*h*eta(M)*F(M)-2*eta(M)+2*eta(M-1))/hˆ2-...
((eta(M-1)+2*h*F(M)*eta(M))*F(M+1)-eta(M-1)*F(M-1))/(2*h)); % gives Eq. (25)
z(2*M)=D2*((2*h*sigma(M)*G(M)-2*sigma(M)+2*sigma(M-1))/hˆ2-...
((sigma(M-1)+2*h*G(M)*sigma(M))*G(M+1)-sigma(M-1)*G(M-1))/(2*h)); % gives
Eq. (28)
z(3*M)=D3*((2*h*m(M)*H(M)-2*m(M)+2*m(M-1))/hˆ2-...
((m(M-1)+2*h*H(M)*m(M))*H(M+1)-m(M-1)*H(M-1))/(2*h)); % gives Eq. (31)
z=z’;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clc;clear all; global h M;
h=1/(M-1); %M=31,45
tf=751;% final time
x=linspace(0,1,M);
tspan=0:1:751;
for i=1:M;
eta0(i)=1;sigma0(i)=1;m0(i)=1;
end
 S. PAMUK, I. ATAC: THE METHOD OF LINES... 195

init=[eta0 sigma0 m0];

init=init’;
[t,y]=ode45(’fparabolic’,tspan,init); % solver that is built up in Matlab
eta=y(:,1:M); sigma=y(:,M+1:2*M); m=y(:,2*M+1:3*M);
figure; % gives Fig.4
plot(x,eta(11:100:750,:)); xlabel(’x’); ylabel(’η’);
figure; % gives Fig.5
plot(x,sigma(11:100:750,:));axis([0 1 0.98 1.007]); xlabel(’x’); ylabel(’σ’);
figure; % gives Fig.6
plot(x,m(11:100:750,:));
xlabel(’x’); ylabel(’m’)
figure; % gives Fig.7
tt=[585 590 674 699];
for i=1:4
subplot(2,2,i); plot(x,sigma(tt(i),:));axis([0 1 0.98 1.007]); xlabel(’x’); ylabel(’σ’);
title([’t=’,num2str(tt(i)-1)]);
end

7. Appendix B
Matlab codes for the construction of the stability matrix S
%%%%%%%%%%%%%%%% S matrix%%%%%%%%%%%%%%%%%%%%
% All of the constants and functions in the following code are the same as in chapter
5.1.
S=zeros(3*M,3*M);
S(1,1)=-D1*(2*(1+h*F(1))/hˆ2+F0*F(1));
S(1,2)=D1*(2/hˆ2-F(2)/(2*h)+F0/(2*h));
for i=2:M-1
S(i,i-1)=D1*(1/hˆ2+F(i-1)/(2*h));
S(i,i)=-2*D1/hˆ2;
S(i,i+1)=D1*(1/hˆ2-F(i+1)/(2*h));
end
S(M,M-1)=D1*(2/hˆ2+F(M-1)/(2*h)-F(M+1)/(2*h));
S(M,M)=D1*((2*h*F(M)-2)/hˆ2-F(M)*F(M+1));
S(M+1,M+1)=-D2*(2*(1+h*G(1))/hˆ2+G0*G(1));
S(M+1,M+2)=D2*(2/hˆ2-G(2)/(2*h)+G0/(2*h));
for i=2:M-1
S(M+i,M+i-1)=D2*(1/hˆ2+G(i-1)/(2*h));
S(M+i,M+i)=-2*D2/hˆ2;
S(M+i,M+i-1)=D2*(1/hˆ2-G(i+1)/(2*h));
end
S(2*M,2*M-1)=D2*(2/hˆ2+G(M-1)/(2*h)-G(M+1)/(2*h));
S(2*M,2*M)=D2*((2*h*G(M)-2)/hˆ2-G(M)*G(M+1));
S(2*M+1,2*M+1)=-D3*(2*(1+h*H(1))/hˆ2+H0*H(1));
S(2*M+1,2*M+2)=D3*(2/hˆ2-H(2)/(2*h)+H0/(2*h));
for i=2:M-1
S(2*M+i,2*M+i-1)=D3*(1/hˆ2+H(i-1)/(2*h));
S(2*M+i,2*M+i)=-2*D3/hˆ2;
S(2*M+i,2*M+i+1)=D3*(1/hˆ2-H(i+1)/(2*h));
end
196 TWMS J. APP. ENG. MATH. V.3, N.2, 2013

S(3*M,3*M-1)=D3*(2/hˆ2+H(M-1)/(2*h)-H(M+1)/(2*h));
S(3*M,3*M)=D3*((2*h*H(M)-2)/hˆ2-H(M)*H(M+1));
S; %The Matrix S
D=real(eig(S)); %The Real Parts of the Matrix S

References
[1] Ahmad, I. and Berzins, M., (2001), MOL solvers for hyperbolic PDEs with source terms, Math.
Comput. Simulat., 56 (2), 115-125.
[2] Erdem, A. and Pamuk, S., (2007), The method of lines for the numerical solution of a mathematical
model for capillary formation: The role of tumor angiogenic factor in the extra-cellular matrix, Appl.
Math. Comput., 186, 891-897.
[3] Korn, G.A., (1999), Interactive solution of partial differential equations by the Method-of-lines, Math.
Comput. Simulat., 49 (1-2), 129-138.
[4] Levine, H.A., Sleeman, B.D. and Nilsen-Hamilton, M., (2000), A mathematical model for the roles
of pericytes and macrophages in the initiation of angiogenesis. I.The role of protease inhibitors in
preventing angiogenesis, Math. Biosc., 168, 77-115.
[5] Mikhail, M.N., (1987), On the validity and stability of the method of lines for the solution of partial
differential equations, Appl. Math. Comput., 22 (2-3), 89-98.
[6] Pamuk, S. and Erdem, A., (2007), The method of lines for the numerical solution of a mathematical
model for capillary formation: The role of endothelial cells in the capillary, Appl. Math. Comput.,
186, 831-835.
[7] Pamuk, S., (2003), Qualitative analysis of a mathematical model for capillary formation in tumor
angiogenesis, Math. Models and Methods Apll. Sci., 13 (1), 19-33.
[8] Schor, A.M., Canfield, A.E., Sutton,A.B., Allen,T.D., Sloan,P., Schor, S.L., Steiner,R., Weisz, P.B.
and Langer,R., (1992), The behavior of pericytes in vitro: relevance to angiogenesis and differentiation,
Angiogenesis: Key Principles-Science-Technology-Medicine, Birkhauser, Basel.
[9] Shakeri, F. and Dehghan, M., (2008), The method of lines for solution of the one-dimensional wave
equation subject to an integral conservation condition, Comput. and Math. with Appl., 56, 2175-2188.
[10] Sharaf, A.A. and Bakodah, H.O., (2005), A good spatial discretisation in the method of lines, Appl.
Math. Comput., 171 (2), 1253-1263.
[11] Younes,A., Konz,M., Fahs, M., Zidane, A. and Huggenberger, P., (2011), Modelling variable den-
sity flow problems in heterogeneous porous media using the method of lines and advanced spatial
discretization methods, Math. Comput. Simulat., 81, 2346-2355.
[12] Teschl, G., (1991), Ordinary Differential Equations and Dynamical Systems, American Mathematical
Society.

Serdal Pamuk obtained his BSc in mathematics from Istanbul University

in 1988. After that he became a research assistant at Marmara University,
where he got a master’s degree in 1992. In 1993, he was granted a scholarship
by YOK to go to overseas to complete his PhD. In 1996, he received his second
MSc from University of Nebraska at Lincoln. After that he completed his
PhD at Iowa State University, USA. Since 2000, he has been working at
Kocaeli University where he became a Professor in 2010.
S. PAMUK, I. ATAC: THE METHOD OF LINES... 197

Irem Atac graduated from Department of Mathematics at Kocaeli Univer-

sity in 2009. At the same year, she began to master education at Kocaeli
University in the program of Mathematics. She was granted a master’s de-
gree in 2012. She continued as a PhD student at the same program. Since
2010, she has been a research assistant in the Faculty of Science and Arts.
Her research interests are applied mathematics, numerical solutions of PDEs
and mathematical biology.