• 沒有找到結果。

酵母菌與爪蟾屬青蛙蛋之基因調控網路:改良型基因演算法

N/A
N/A
Protected

Academic year: 2021

Share "酵母菌與爪蟾屬青蛙蛋之基因調控網路:改良型基因演算法"

Copied!
43
0
0

加載中.... (立即查看全文)

全文

(1)

୯ ҥ Ҭ ೯ ε Ꮲ

ႝᐒᆶ௓ڋπำس

ᅺγፕЎ

ሇ҆๵ᆶПᛶឦߙຂೈϐ୷Ӣፓ௓ᆛၡǺ

ׯؼࠠ୷Ӣᄽᆉݤ

Genetic Regulatory Network of Yeast / Xenopus Frog Egg

Ǻ

Improved Genetic Algorithm

ࣴ ز ғǺڬ ৎ ፣

ࡰᏤ௲௤Ǻ׵ ઒ బ റγ

ֆ ۩ ࣔ റγ

(2)

ሇ҆๵ᆶПᛶឦߙຂೈϐ୷Ӣፓ௓ᆛၡǺׯؼࠠ୷Ӣᄽᆉݤ

Genetic Regulatory Network of Yeast / Xenopus Frog EggǺ

Improved Genetic Algorithm

ࣴ ز ғǺڬৎ፣

ࡰᏤ௲௤Ǻ׵઒బ റγ

ֆ۩ࣔ റγ

Student

ǺChia-Hsien Chou

AdvisorǺDr. Tsu-Tian Lee

Dr. Shinq-Jen Wu

୯ ҥ Ҭ ೯ ε Ꮲ ႝᐒᆶ௓ڋπำس

ᅺγፕЎ

A Thesis

Submitted to Department of Electrical and Control Engineering College of Electrical and Computer Engineering

National Chiao Tung University in partial Fulfillment of the Requirements

for the Degree of Master

in

Electrical and Control Engineering July 2006

Hsinchu, Taiwan, Republic of China

(3)

ሇ҆๵ᆶПᛶឦߙຂೈϐ୷Ӣፓ௓ᆛၡǺׯؼࠠ୷Ӣᄽᆉݤ

ᏢғǺڬৎ፣

ࡰᏤ௲௤Ǻ׵઒బ റγ

ֆ۩ࣔ റγ

୯ҥҬ೯εᏢႝᐒᆶ௓ڋπำسȐࣴز܌ȑᅺγ੤

ճҔׯؼࠠ୷ӢᄽᆉݤǴଞჹሇ҆๵ᆶПᛶឦߙຂೈϩձ଺୷Ӣፓ௓ᆛ ၡ modified power-law model ᆶ S-system ࡌኳǶᙖҗሇ҆๵ჴᡍکПᛶឦ

Michaelis-Menten model܌ளډޑਔ໔ᗺၗ਑໣଺૽ግǴаᕇڗന٫ϯޑୖ

ኧǴӢࣁ modified power-law model ک S-system ёаమཱӦඔॊ୷Ӣӧғԋ ܈੃઻ϸᔈਔࢂ໽ϯբҔᗋࢂ׭ڋբҔǴ᠙ܭٿޣޑ੝ՅǴჹሇ҆๵ޑಒ झຼයᆶӧПᛶឦߙຂೈಒझຼයύޑԖํϩ຋௓ڋǴૈܴᕕځ୷Ӣޑϸ ᔈ௃׎ǶڀԖᎂ౽բҔޑׯؼࠠ୷Ӣᄽᆉݤόՠёа଺ډӄୱཛྷ൨Ǵᗋё аᆒमЬကޑཷۺڗளന٫ޑঁᡏǶനಖளډޑ୷Ӣፓ௓ᆛၡёаගٮ๏ ғނᏢৎӧሇ҆๵کПᛶឦߙຂೈಒझຼයБय़଺׳ుΕޑࣴزǶ

(4)

Genetic Regulatory Network of Yeast / Xenopus Frog EggǺ

Improved Genetic Algorithm

Student

ǺChia-Hsien Chou

Advisor

ǺDr. Tsu-Tian Lee

Dr. Shinq-Jen Wu

Department (Institute) of Electrical and Control Engineering

National Chiao Tung University

Abstract

An improved genetic algorithm is proposed to achieve gene regulatory network modeling of Xenopus frog egg in S-system and yeast in modified power-law model respectively. Via the time-course datasets from experiment of yeast and Michaelis-Menten model of Xenopus, the optimal parameters are learned. The modified power-law model and S-system can clearly describe activative and inhibitory interaction between genes as generating and consuming process. We concern cell cycle of yeast and the mitotic control in cell cycle of Xenopus frog egg to realize gene reactions. The proposed improved genetic algorithm can achieve global search with migration and keep the best individual with elitism operation. The generated gene regulatory networks can provide biological researchers for further experiments in yeast and Xenopus frog egg cell cycle.

(5)

Acknowledgements

२Ӄךाགᖴ׵઒బਠߏکֆ۩ࣔԴৣǴӧ೭ٿԃ๏ךޑፌፌ௲ᇧǴӧ ࣴزБय़όսெӦ௲௤ך΋٤ᢀۺၟБݤǴ٬ךᕇ੻ؼӭǴᏢғόയགᐟǴ Ψࡐགᖴύ҅εᏢϯπسޑЦ೹౰௲௤ᆶ࢒ᇞᓪᏢߏ๏ϒךჹғނၗૻࣴ زޑ΋٤ࡰ௲ǴᗋԖֆࡹၰᏢߏၟך΋ଆவ٣୯ࣽ཮ीฝࣴزǴΨ๏ϒךࡐ ӭޑႴᓰᆶ௲ᏤǴߚதགᖴдॺǴڐշךٿԃޑࣴزғఱǶ ӧፐ཰΢Ǵჴᡍ࠻ޑᏢߏЦ╨ᗶǴ஭௘Ǵ݅ਤӹǴ݅ࢩ᎛ǴֆܲڻǴֆ ۘ᎔аϷ৪⊭ਤᔅշךှ،ӧঅፐਔ܌ၶډޑୢᚒǴᗋԖځдჴᡍ࠻ޑӕᏢ ݅ࡏֻǴ஭ࡿሎǴ஭मቺӧঅፐਔεৎϕ࣬૸ፕᔅԆǴΜϩགᖴǶԜѦǴ׳ Ԗᒘ೚ӭܻ϶ޑᆒઓЍ࡭Ǵࣗ܈ჴ፦ڐշǴךωૈ໩ճֹԋᏢ཰Ǵࡸ΋ᅅ࿤Ǵ ёૈۘԖ΋٤҂ගϷޑܻ϶ॺǴӧԜ֡΋ٳठᖴǶ നࡕाགᖴךޑР҆ڬᓪϡᆶ࢒म໦Ǵךޑۊۊڬ܃ۇǴךޑ׌׌ڬৎ ѶǴᗋԖךޑζ϶ጰۏ։ǴऩόࢂдॺߏΦаٰޑЍ࡭ǴόёૈԖךϞϺޑ λλԋ݀Ƕ

(6)

Contents

ᄔा ...i

Abstract ...ii

Acknowledgments ...iii

Contents ...iv

List of tables ...vi

List of figures ...vii

1 Introduction ...1

1.1 Research Background ...1

1.2 Literature Discussion ...1

1.3 Content Organization ...3

2 Biological Systems ...4

2.1 Yeast cell cycle ...4

2.2 M phase control of Xenopus frog egg ...6

3 Improved Genetic Algorithm ...10

3.1 Introduction ...10

3.2 IEDO ...11

3.3 Reproduction ...12

3.4 Crossover ...13

3.5 Probability Mutation and Elitism ...14

3.6 Migration ...15

3.7 Fitness ...16

4 Modeling and Simulation Results ...18

(7)

4.2 M phase control of Xenopus frog egg ...21

5 Conclusion ...30

References ...31

(8)

List of tables

(9)

List of figures

2.1 The four phases of cell cycle ...4

2.2 The concentrations of cyclin, unphosphorylated cyclin-Cdc2 and Tyr-15 phosphorylated cyclin-Cdc2 ...8

2.3 The concentrations of doubly phosphorylated cyclin-Cdc2 and Thr-161 phosphorylated cyclin-Cdc2 ...8

2.4 The concentrations of Cdc25 and IE ...9

2.5 The concentrations of Wee1 and APC ...9

3.1 Flow chart for improved genetic algorithm ...10

3.2 Population definition ...11

3.3 Rule of reproduction ...13

3.4 Crossover operation ...14

3.5 Probability mutation operation ...15

4.1 The gene regulatory network of the generated modified power-low system ...20 4.2 Training dataset-1 ...21 4.3 Training dataset-2 ...22 4.4 Training dataset-3 ...22 4.5 Training dataset-4 ...23 4.6 Cyclin evolution ...25

4.7 Unphosphorylated cyclin-Cdc2 evolution ...25

(10)

4.9 Doubly phosphorylated cyclin-Cdc2 evolution ...26

4.10 Thr-161 phosphorylated cyclin-Cdc2 evolution ...27

4.11 Cdc25 evolution ...27

4.12 Wee1 evolution ...28

4.13 IE evolution ...28

(11)

Chapter 1

Introduction

1.1 Research Background

As the rapid development in cDNA microarray technologies, time-course gene dataset becomes available day by day. Hence, the construction of gene regulatory networks and signal transduction cascades for complicated biological systems has come of age. In order to approximate biological behavior for controlling metabolic/biological reaction, more and more experiments are set up for achievement of quantitative control. After post-genomic era, new scientific and technological methods on the biotechnology such as microarray technology are developed to bring massive biological knowledgeable dataset. Now system biologists are trying to describe biochemical phenomenon via mathematical model. With the mathematical model, we can realize the detailed genes-genes interaction, simulate the gene regulatory network and predict gene behavior.

1.2 Literature Discussion

Numerous models are proposed to describe the gene network such as Boolean network, Bayesian network, Michaelis-Menten model, and S-system. Boolean network is to reconstruct gene regulatory network via Boolean function and

(12)

express gene relationship in graphical way [1, 2], which distinguish gene states to be INPUT and OUTPUT. At any time points, the state values of chosen INPUT genes are set to be 1 and 0 for non-chosen genes; the states values are given in the similar way as OUTPUT. Further, Bayesian network can also the probabilistic relationships of genes [2, 3]; joint probability distributions among genes are calculated to construct the graphical model. Michaelis-Menten model is nonlinear differential equations to describe the metabolic concentration in the biological system [4]. S-system is another nonlinear differential type expressed in power-law formalism [5, 6]. S-system describes gene regulation not only in mathematical description but also can further express into graphical form to show the activatory and inhibitory operation directly. Each equation is composed by synthesis and degradation flux; and the activation and inhibition relationship are shown in positive and negative kinetic order, respectively. In this paper, we shall develop the general S-system and another reformed nonlinear differential system, which is a modified power-low model modified from equation in [7, 8], to find out the gene regulatory network of yeast cell cycle and Xenopus cell cycle M phase control from microarray dataset, respectively.

However, the construction of such a highly nonlinear equation is a tough work. Chen and the authors first reform the nonlinear differential equation into linear form and then resolve it via linear algebra [7]. In these years, some researchers are devoted to infer gene regulatory network with various intelligent computation technologies such as hybrid differential evolution, genetic algorithm, genetic programming, ..., etc. Wang use Hybrid differential evolution and genetic algorithm to obtain the global optimal solution for highly nonlinear system and

(13)

various biochemical system [9, 10]. Kikuchi and coauthors use a genetic algorithm to transform parameters into individuals first and solve optimal parameters via evolution procedure [11]. Sakamoto and coauthors use genetic programming to develop the gene regulatory network in a tree form [12]. In this work, we shall adopt improved genetic algorithm [13] to infer the gene regulatory networks of yeast cell cycle in modified power-law model and Xenopus frog egg cell cycle in S-system. Improved evolutionary direction operator (IEDO), migration operation and elitism are combined into genetic program for global optimal, fast and best-optional searching. The input/output datasets, yeast cell cycle dataset [14], generated from Michaelis-Menten model of mitotic control in Xenopus frog eggs [15, 16], are used to train the genetic networks for searching the optimal parameters of the corresponding modified power-law model and S-system, respectively.

1.3 Content Organization

This paper is organized as follows: the biological systems, yeast cell cycle and cell cycle M phase control model of the Xenopus frog egg, are described in Chapter 2. Improved genetic algorithm is shown in Chapter 3. Chapter 4 shows the modeling and simulation results. Chapter 5 is the conclusion.

(14)

Chapter 2

Biological System

2.1 Yeast cell cycle

Research in cell cycle is very important not only for realizing cell reproduction but also for realizing cancer development. Figure 2.1 is the cell cycle that includes four phases (G1ÆSÆG2ÆM). Cell grows up in G1 phase, produces RNA and synthesizes protein. During S phase, DNA is duplicated to produce two similar daughter cells. The cell continues to grow and produce protein and prepares to enter M phase during G2 phase. As DNA replication is completed, the cell enters M phase and divides.

(15)

Table 2.1 The genes of the Yeast cell cycle.

Gene Description

CDC28 Catalytic subunit of the main cell cycle cyclin-dependent kinase CLN3

role in cell cycle START; involved in G(sub)1 size control;

G1/S-specific cyclin, interacts with Cdc28p protein kinase to control events at START

SWI4

Involved in cell cycle dependent gene expression; both Swi4p and Swi6p are required for the in vivo protection of the SCB sequences at any cell cycle stage

SWI6 Involved in cell cycle dependent gene expression MBP1 transcription factor

FUS3 Required for the arrest of cells in G(sub)1 in response to pheromone and cell fusion during conjugation

FAR1 Inhibitor of Cdc28p/Cln1p and Cdc28p/Cln2p complexes involved in cell cycle arrest for mating; Factor arrest protein

CLN1 G1 cyclin; role in cell cycle START CLN2 G1 cyclin; role in cell cycle START

SIC1 P40 inhibitor of Cdc28p-Clb5 protein kinase complex CLB5

role in DNA replication during S phase; additional functional role in formation of mitotic spindles along with Clb3 and Clb4; B-type cyclin involved in S-phase initiation

CLB6 role in DNA replication during S phase; B-type cyclin involved in S-phase initiation

CDC6

Protein involved in initiation of DNA replication; Protein that regulates initiation of DNA replication through binding to origins of replication at the end of mitosis, directing the assembly of MCM proteins and the pre-replication complex

CDC20 Cell Division Cycle; Required for onset of anaphase; adaptor for APC GRR1 F box protein with several leucine rich repeats

CDC4

Init. of DNA synthesis & spindle pole body separation; dispensable for both mitotic and meiotic spindle pole body dupl.; essential for mitotic but not premeiotic DNA synth.; wt levels of synaptonemal complexes and intragenic recombination

(16)

The yeast cell cycle gene expression data is collected by Spellman [14]. The dataset were covered six experimental conditions (CLN2; CLN3; ALPH, CDC15, CDC28 and ELU). We use the gene time-course data from experimental condition CDC28, which contains 24 time points. We concentrate in G1 and S phase and set up our dataset from the sub network of yeast cell cycle pathway. The descriptions and datasets are available at http://cellcycle-www.stanford.edu. This dataset in Table 2.1 involves 16 genes.

2.2 M phase control of Xenopus frog egg

Michaelis-Menten model are concerned to describe the mitotic control in cell-cycle of Xenopus frog egg [15, 16],

1 1 2 1 3 1 2 5 2 2 25 3 3 1 3 2 25 2 3 4 , (1) , (2) , pp wee cak wee cak pp x k k x k x x k x k k k x k x k x x k x k k k x k x              

4 5 25 2 4 3 5 2 2 5 25 4 5 6 6 6 6 6 (3) , (4) , (5) 1 , 1 wee pp cak cak pp wee a b a b x k x k k k x k x x k x k k k x k x k x x k x x K x K x                

7 5 7 7 7 7 5 8 8 8 8 8 9 (6) 1 , (7) 1 1 , (8) 1 f e e f g h g h k x k x x x K x K x k x x k x x K x K x k x              8

9

9 9 9 1 , (9) 1 c d c d x x k x K x K x     

(17)

with

2 2 9 2 2 7 25 25 6 25 25 , , ,

wee wee wee wee

k V x V V k V x V V k V x V V c cc c cc  c  cc c  cc  c

where xi, i = 1, 2, ···, 9, are the concentrations or activities of cyclin,

unphosphorylated cyclin-Cdc2, Tyr-15 phosphorylated cyclin-Cdc2, doubly phosphorylated cyclin-Cdc2, Thr-161 phosphorylated cyclin-Cdc2 activated by four enzymes Cdc25, Wee1, IE and APC, respectively. Eqs. (1) ~ (4) describe four phosphorylation states of the cyclin-Cdc2 dimer. The Thr-161 phosphorylated cyclin-Cdc2 represents the M phase promoting factor (MPF). The concentration of MPF can control the phases during cell cycle. For instance, high concentration of MPF can make the cell to divide to two child cells.

The numerical solution of Michaelis-Menten model is shown as Figure 2.2 ~ 2.5. The parameter values Ka=0.1, Kb=1.0, Kc=0.01, Kd=1.0, Ke=0.1, Kf=1.0,

Kg=0.01, Kh=0.01, ka=2.0, kb=0.1, kc=0.13, kd=0.13, ke=2.0, kf=0.1, kg=2.0, kh=0.15, k1=0.01, k3=0.5, kcak=0.64, kpp=0.004, V25c 0.017 , , , , , 25 0.17 Vcc 2 0.005

(18)

0 20 40 60 80 100 120 140 160 180 200 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 time [min] c o nc en tr a ti o n Cyclin Unphosphorylated cyclin-Cdc2

Tyr-15 phosphorylated cyclin-Cdc2

Figure 2.2 The concentrations of cyclin, unphosphorylated cyclin-Cdc2 and

Tyr-15 phosphorylated cyclin-Cdc2.

0 20 40 60 80 100 120 140 160 180 200 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 time [min] c o nc en tr a ti o n Thr-161 phosphorylated cyclin-Cdc2

doubly phosphorylated cyclin-Cdc2

Figure 2.3 The concentrations of doubly phosphorylated cyclin-Cdc2 and

(19)

0 20 40 60 80 100 120 140 160 180 200 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 time [min] c o nc en tr a ti o n Cdc25 IE

Figure 2.4 The concentrations of Cdc25 and IE.

0 20 40 60 80 100 120 140 160 180 200 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 time [min] c o nc en tr a ti o n Wee1 APC

(20)

Chapter 3

Improved Genetic Algorithm

3.1 Introduction

Based on general genetic algorithm, improved genetic algorithm includes improved evolutionary direction operator to speed the searching; migration operator to escape from bogging down into local solution and elitism to keeping the best to be passed down always. The flow chart is shown in Figure 3.1.

(21)

The algorithm initializes a population first. The definition of population is shown in Figure 3.2. The chromosome in individual represents parameter. And then, we evaluate the fitness value of each individual in the population and keep the better individuals and eliminates the worse individuals though evolution procedures.

Figure 3.2 Population definition.

3.2 IEDO

Three preferred fitness value and its associated individuals are chosen to decide the evolutionary direction in the population. Improved evolutionary direction operator (IEDO) has the ability for both local and global search synchronously.

(22)

optimal solution. Besides, the IEDO operator can avoid bogging down in the local optimal solution, which conventional GA is usually stuck into.

After completing all fitness values of individuals, we choose preferred fitness values denote Fb, Fs, Ft and its associated individuals denote Ib, Is, It. And

then, the new individual denote Iideo is calculated as

(10) (11) s t 1 1 2 2 iedo b b b max min iedo iedo I = I + r * D * ( I - I ) + r * D * ( I - I ), I = max(min(I ,I ),I ),

where r1 and r2 are two random numbers, r1, r[0, 1]; D1 and D2 are the

magnitude of two evolutionary directions to be 1; Imax and Imin are the upper and

lower bound, respectively. The new fitness value Fnew of the Iideo is calculated; if

the Fnew is better than one of the three preferred fitness values, it would be

replaced that.

3.3 Reproduction

The probability of reproduction directly depends on the fitness of the individuals. The individual with better fitness has high probability of reproduction. In contrast, the individual with worse fitness has low probability of reproduction. The rule of reproduction is shown in Figure 3.3.

(23)

Figure 3.3 Rule of reproduction.

3.4 Crossover

Two-point crossover operation shown in Figure 3.4 is adopted and operates according to crossover probability, which involves selection of two crossover cut-points randomly and then exchanges the chosen two cut-points genes of parent individuals to generate two child individuals. And further, randomly select one of the new individuals to replace father or mother individuals.

(24)

Figure 3.4 Crossover operation.

3.5 Probability Mutation and Elitism

Different from the conventional GA to choose only one gene in individual randomly for mutation operation, all genes in individuals are chosen and their mutation probability are assigned by the designer for exchanging their original values in Figure 3.5. This will bring excessive diversity in population and hence may fail to converge to temperately optimal solution. Therefore, we adopt elitism operator to decrease this effect. Elitism operator is to keep the best individual to survive for each generation and hence to ensure good characteristic to pass down

(25)

always.

Figure 3.5 Probability mutation operation.

3.5 Migration

To wider the search space, a migration operator is done to get a new and diverse population. The degree of population diversity Ș is to check if the migration should be performed. 2 _ 1 1 1 0, if , (12) 1, otherwise , (13) _ ( 1) ij bj ij bj Dim I NP ij i j x x temp x temp Dim I NP

H

K

 °° ® ° °¯   u 

¦ ¦

­

(26)

respectively the j-th chromosome in the i-th individual and the best individual; NP is the number of individual; Dim I is the dimension of individual; Ș is in the range between 0 and 1. İ1[0,1] is a tolerance-threshold of population diversity

for migration; if Ș is small than İ1, migration operate to generate a new

chromosome as follows.

,min 2 ,m 1 ,m j j x in ,max ,min ax 2 , if , (14) , ortherwise bj j bj bj j j ij bj bj x x x r x x r x x x x r x ­ ° ° ® ° °¯   u  !   u 

where xj,max and xj,min are the upper and lower bound of the j-th chromosome,

respectively. The r1 and r2 are two random numbers, r1, r2[0, 1].

3.6 Fitness

Every individual is evaluated by its fitness value, which keeps the better individuals and eliminates the worse individuals. We adopt two different methods to evaluate fitness. The fitness of an individual in yeast cell cycle is defined as

0 0

2 1 1 , (15) ei i i j fitness X t j t X t j t kinetic order  '  ' 

¦¦

¦

-1 n N

where the Xi(t) is approximate value of the i-th variable at time t, Xei(t) is original

(27)

The fitness of an individual in Xenopus M phase control is defined as

_ -1 Dim I N 2 0 0 1 1 -, (16) _ ei i i j X t j t X t j t fitness Dim I N  '  ' u

¦ ¦

where N is number of time points; Xei(t) and Xi(t) are experiment value and

(28)

Chapter 4

Modeling and Simulation Results

4.1 Yeast cell cycle

The mathematical model in yeast cell cycle is approximated from the model adopted in [7, 8].

( ) ( ) ( ), =1,2, , , (17)

i i i i

X t G t O X t i  n

where Gi(t) is the transcription rate, Ȝi is the self-degradation rate and n is the

number of the variable, Xi(t) is the concentration of the i-th gene at time t. Gi(t) is

a nonlinear function,

^

`

1 ( ) , (18) 1 exp , , i ij j j G t a u t

D

ª«

E G

J

º» ¬ ¼   

¦

m 1

0 ( , , ) . (19) ( ) 0 j j j j j j j j j t t t u t t t t

E

G

E

G

E

G

E

G

E G

E

G

E

E

G

G

E

G

­ ° ° ° ° ® ° ° ° °¯ d     d d   d d  t 

We use a power-law function Vi(t) to approximate the nonlinear function Gi(t) in

(29)

1 ( ) fij( ), (20) i i j j V t O

–

n X t k

where Ȝi is the rate constant and fij is kinetic order. Further, the degradation term

in Eq. (17) is replaced by ȖiXii( )t to emphasize how a gene reacts itself. The

kinetic orders, fij and Ȗi, can be positive or negative; positive kinetic orders

indicate activating influences, but negative kinetic orders mean inhibition. In other words, the following modified power-law dynamic model is proposed.

1 ( ) ( , ) ( ) ( ) ( ) ( ), =1,2, , , (21) i ij i i i i i i n f k i j i i j X t f V t X t X t X t i n

J

O

J

 

–

X P  k 

where n is the number of the variables; the vector X in Eq. (21) indicates all genes in the yeast cell cycle; the vector P in Eq. (21) consists the rate constants, Ȝi and Ȗi, and kinetic orders, fij and ki. According to 16 genes of yeast cell cycle,

there are 16 differential equations with 288 parameters.

Figure 4.1 is the pathway for the modified power-low model whose fitness is 1.1104688E-02. Black lines represent activation reaction and red lines represent inhibition reaction. The start point of the lines is the reactant and the end point is the product. For instance, the concentration of CDC28 increases rapidly as the concentrations of MBP1, CLN1, CDC6, CDC20 and GRR1 increase; however, the concentration of CDC28 decreases rapidly as the concentrations of CLN3, SWI4, FUS3, FAR1, CDC4, CLB6 andCLB6 increase.

(30)

Figure 4.1 The gene regulatory network of the generated modified power-low

(31)

4.2 M phase control of Xenopus frog egg

We shall generate the training dataset from the Eqs. (1) ~ (9) to generate the corresponding S-system model of the frog cell cycle to further realize the gene-gene inhibitory and activatory operation for gene and enzyme synthesis and decomposition. The training dataset is shown in Figure 4.2 ~ 4.5.

0 10 20 30 40 50 60 70 80 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 time [min] c o nc en tr at io n Cyclin Unphosphorylated cyclin-Cdc2

Tyr-15 phosphorylated cyclin-Cdc2

(32)

0 10 20 30 40 50 60 70 80 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 time [min] c o nc en tr at io n Thr-161 phosphorylated cyclin-Cdc2

doubly phosphorylated cyclin-Cdc2

Figure 4.3 Training dataset-2.

0 10 20 30 40 50 60 70 80 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 time [min] c o nc en tr at io n Cdc25 IE

(33)

0 10 20 30 40 50 60 70 80 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 time [min] c o nc en tr at io n Wee1 APC

Figure 4.5 Training dataset-4.

S-system use power-law flux to describe the synergism and saturation of the biological system, 1 1 , for 1,2,..., , (22) ij ij g h i i j i j j j x

D

–

n x 

E

–

n x i n

where xi is the state variable or reactant; n is the number of xi.Įi is the production

rate-constant and ȕi is the degradation rate-constant; both can be positive or zero.

gij and hij, are kinetic orders; their values can be positive to indicate activating

influences or negative to denote inhibition. We now construct our S-system structure for Xenopus frog egg as

(34)

19 19 11 g 11 h (23) g h x

D

x x 

E

x x 23 25 26 27 29 23 26 27 29 21 22 22 32 33 34 36 37 1 1 1 9 1 1 9 5 7 7 2 2 1 2 3 6 9 2 2 3 6 9 7 3 3 2 3 4 6 9 (24) g g g g g h h h h g g h g g g g g x x x x x x x x x x x x x x x x x x x x

D

E

D

   39 32 33 36 37 39 43 44 45 46 47 49 44 45 46 47 49 52 54 55 56 57 59 54 55 56 57 59 7 3 2 3 6 9 5 7 5 7 4 4 3 4 6 9 4 4 6 9 5 5 2 4 5 6 7 9 5 4 5 6 7 9 (25) (26) g h h h h h g g g g g g h h h h h g g g g g g h h h h h x x x x x x x x x x x x x x x x x x x x x x x x x x x x x

E

D

E

D

E

     65 66 65 66 75 77 75 77 85 88 5 5 6 6 6 6 6 7 7 5 7 7 5 7 5 8 8 8 8 (27) (28) (29) g g h h g g h h g g x x x x x x x x x x x x x

D

E

D

E

D

E

      85 88 98 99 98 99 5 8 9 9 8 9 9 8 9 (30) (31) h h g g h h x x x

D

x x 

E

x x

Note that since concentrations of x1, x2, x3, x4and x5 are too small as compared to

other variables. According to Eqs. (23) ~ (31), there are 83 parameters to estimate. The scale-up operation is adopted to normalize all states variables to a computation reasonable range to improve the computation error. Another test data from Michaelis-Menten model is used to demonstrate the performance of the improved genetic algorithm program, Figure 4.6 ~ 4.14 is the simulation results with the estimated fitness 3.0766122E-08 for N=80,000, Dim_I=83. The low fitness value ensures the good fitting of the simulation results with the datasets and also guarantees the reliability of the generated S-system. From the constructed S model, we can realize the interaction between various genes in Xenopus frog egg. For instance, the concentration of x2 increases rapidly as

concentrations of x1, x3 and x5 increase; the concentration of x2 decreases rapidly

(35)

0 20 40 60 80 100 120 140 160 180 200 0.013 0.014 0.015 0.016 0.017 0.018 0.019 0.02 x1 (Cyclin) time [min] c o nc en tr at io n Test data S-system

Figure 4.6 Cyclin evolution.

0 20 40 60 80 100 120 140 160 180 200 0.006 0.007 0.008 0.009 0.01 0.011 0.012 0.013 0.014 0.015 x2 (Unphosphorylated cyclin-Cdc2) time [min] c o nc en tr at io n Test data S-system

(36)

0 20 40 60 80 100 120 140 160 180 200 0 1 2 3 4 5 6 7x 10

-3 x3 (Tyr-15 phosphorylated cyclin-Cdc2)

time [min] c o nc en tr at io n Test data S-system

Figure 4.8 Tyr-15 phosphorylated cyclin-Cdc2 evolution.

0 20 40 60 80 100 120 140 160 180 200 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14

x4 (doubly phosphorylated cyclin-Cdc2)

time [min] c o nc en tr at io n Test data S-system

(37)

0 20 40 60 80 100 120 140 160 180 200 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 x5 (Thr-161 phosphorylated cyclin-Cdc2) time [min] c o nc en tr at io n Test data S-system

Figure 4.10 Thr-161 phosphorylated cyclin-Cdc2 evolution.

0 20 40 60 80 100 120 140 160 180 200 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 x6 (Cdc25) time [min] c o nc en tr at io n Test data S-system Figure 4.11 Cdc25 evolution.

(38)

0 20 40 60 80 100 120 140 160 180 200 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 x7 (Wee1) time [min] c o nc en tr at io n Test data S-system

Figure 4.12 Wee1 evolution.

0 20 40 60 80 100 120 140 160 180 200 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x8 (IE) time [min] c o nc en tr at io n Test data S-system Figure 4.13 IE evolution.

(39)

0 20 40 60 80 100 120 140 160 180 200 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 x9 (APC) time [min] c o nc en tr at io n Test data S-system

(40)

Chapter 5

Conclusion

Improved genetic algorithm technique, which involves IEDO, migration operation and Elitism, is adopted to construct mathematical models for yeast and Xenopus frog egg. The gene time-course data form experiment of yeast cell cycle is used to construct gene regulatory network in modified power-law model. The training and test dataset are generated from Michaelis-Menten metabolic model of mitotic cell-cycle control of Xenopus frog egg in S-system. The two proposed gene regulatory networks reveal activatory and inhibitory operations for gene/enzyme synthesis and decomposition. Hence, the networks can provide biological researchers for further experiments in yeast and Xenopus frog egg cell cycle.

(41)

References

[1] Akutsu, T., Miyano, S. and Kuhara, S., “Identification of genetic networks from a small number of gene expression patterns under the Boolean network model”, Proceedings of the Pacific Symposium Biocomputation, Vol. 4, p.p. 17–28, 1999.

[2] De Jong, H., “Modeling and simulation of genetic regulatory systems: a literature review”, J. Comput. Biol, Vol. 9, p.p. 67–103, 2002.

[3] Pe’er, D., Regev, A., Elidan, G. and Fridman, N., “Inferring subnetworks from expression profiles”, Bioinformatics, Vol. 17, p.p. 215–224, 2001. [4] Gonick, L. and Wheelis, M., The Cartoon Guide to Genetics, Collins, 1991. [5] Savageau, M.A., Biochemical System Analysis: A Study of Function and

Design in Molecular Biology, Addison-Wesley, 1976.

[6] Voit, E.O., Computational Analysis of Biochemical System, Cambridge University Press, 2000.

[7] Chen, H.C., Lee, H.C., Lin, T.Y., Li, W.H. and Chen, B.S., “Quantitative characterization of the transcriptional regulatory network in the yeast cell cycle”, Bioinformatics, Vol.20, pp. 1914-1927, 2004.

[8] Singer, R. and Penman, S., “Messenger RNA in HeLa cells: kinetics of formation and decay”, J. Mol. Biol., Vol.78, pp.321-328, 1973.

[9] Tsai, K.Y. and Wang, F.S., “Evolutionary Optimization with Data Collocation for Reverse Engineering of Biological Networks” , Bioinformatics, Vol. 21, p.p. 1180–1188, 2005.

[10] Chiou, J. P. and Wang, F. S., “A hybrid method of differential evolutionary with application to optimal control problems of a bioprocess system”, Proceedings 1998 IEEE on Evolutionary Computation Conference, p.p. 627–632, 1998.

(42)

[11] Kikuchi, S., Tominaga, D., Arita, M., Takahashi, K. and Tomita, M., “Dynamic modeling of genetic networks using genetic algorithm and S-system”, Bioinformatics, Vol. 19, p.p. 643–650, 2003.

[12] Sakamoto, E. and Iba, H., “Inferring a System of Differential Equations for a Gene Regulatory Network by using Genetic Programming”, Proceedings of the Evolutionary Computation Congress, Vol. 1, p.p. 720–726, 2001. [13] Chiang, C.L., Su, C.Y. and Wang, F.S., “Augmented Lagrangian Method

for Evolutionary Optimization of Mixed-Integer Nonlinear Constrained Problem”, International Mathematical Journal, Vol. 2, No. 2, p.p. 119–154, 2002.

[14] Spellman, P.T., Sherlock, G., Zhang, M.Q., Iyer, V.R., Anders, K., Eisen, M.B., Brown, P.O., Botstein, D. and Futcher, B., “Comprehensive identification of cell cycle regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization”, Mol. Bio. Cell, Vol. 9, pp. 3273-3297, 1998.

[15] Schmidt, H. and Jacobsen E.W., “Linear systems approach to analysis of complex dynamic behaviors in biochemical networks”, IEE Syst. Biol., Vol. 1, No. 1, p.p. 149–158, 2004.

[16] Marlovits, G., Tyson, C.J., Novak, B. and Tyson, J.J., “Modeling M-phase control in Xenopus oocyte extracts: the surveillance mechanism for unreplicated DNA”, Biophys. Chem., Vol. 72, p.p. 169–184, 1998.

(43)

Vita

ۉ

ӜǺڬৎ፣

рғВයǺ67ԃ10Д20В

ೣǺଯ໢ѱ

ᐕǺଯ໢ѱҥߞက୯λ 74~80

ଯ໢ѱҥϖᅽ୯ύ 80~83

୯ҥଯ໢π୘஑ਠ ႝᐒπำࣽ 83~88

୯ҥѠ᡼ࣽמεᏢ ႝᐒπำس 88~90

୯ҥҬ೯εᏢ ႝᐒᆶ௓ڋπำࣴز܌ 93~95

數據

Figure 2.1 The four phases of cell cycle.
Figure 2.2 The concentrations of cyclin, unphosphorylated cyclin-Cdc2 and  Tyr-15 phosphorylated cyclin-Cdc2
Figure 2.4 The concentrations of Cdc25 and IE.
Figure 3.1 Flow chart for improved genetic algorithm.
+7

參考文獻

相關文件

Programming languages can be used to create programs that control the behavior of a. machine and/or to express algorithms precisely.” -

➢The input code determines the generator output. ➢Understand the meaning of each dimension to control

In the school opening ceremony, the principal announces that she, Miss Shen, t is going to retire early.. There will be a new teacher from

• A function is a piece of program code that accepts input arguments from the caller, and then returns output arguments to the caller.. • In MATLAB, the syntax of functions is

Keywords: pattern classification, FRBCS, fuzzy GBML, fuzzy model, genetic algorithm... 第一章

In the citric acid cycle, how many molecules of FADH are produced per molecule of glucose.. 111; moderate;

4.2 Copy the selected individuals, then apply genetic operators (crossover and mutation) to them to produce new individuals.. 4.3 Select other individuals at random and

本論文之目的,便是以 The Up-to-date Patterns Mining 演算法為基礎以及導 入 WDPA 演算法的平行分散技術,藉由 WDPA