Applications of Genetic Algorithms

1
Applications of Genetic Algorithms

• Namrata Khemka
• CPSC 605
• November 2004

2
Topics

• Traveling Salesman Problem (TSP)
• Protein Folding

3
Traveling Salesman Problem
4
Introduction to TSP
5
NP-Hard

• Number of steps required to solve the problem
  increases at a very high rate
• Iterative Method
• O(n!)

6
Traveling Cities
1
1
1
7
How quickly the numbers grow?
8
Genetic Algorithm Approach
Chromosome
A
B
C
D
E
F
Population
Gene
City
9
Example
10
Initialize the population

• Population Size 5
• EDCBA (127)
• CDAEB (111)
• BDCEA (152)
• ABCED (119)
• EBCAD (110)

11
Step 3 Select 2 best ones

• EDCBA (127)
• CDAEB (111)
• BDCEA (152)
• ABCED (119)
• EBCAD (110)

12
Crossover
C
D
A
B
E
A
B
E
C
D
A
C
D
C
D
C
D
A
B
E
C
D
A
A
B
E
B
E
B
E
E
A
C
D
A
B
E
D
A
C
D
A
A
B
E
D
C
A
B
E
B
E
13
Mutation
E
C
D
A
B
E
A
C
D
A
B
A
B
E
D
C
A
B
E
D
C
14
Example

• http//www-cse.uta.edu/cook/ai1/lectures/applets/
  gatsp/TSP.html

15
Protein FoldingAndGenetic Algorithms
16
Introduction

• Protein functionality depends on the composition
  and conformation.

Alpha-Helices
Beta-Sheets
http//www.spectroscopynow.com/ftp_images/2frag20_
Protein_folding.jpg
17
Introduction Cont..

• Kinetically or thermodynamically controlled (Open
  question)
• Global Energy Minimum
• The Protein Folding Problem
• Ron Unger and John Moult (1992)
• Dr. Graham Cox (2004)
• School of Chemistry, University of Birmingham

18
Lattice Bead Model
P
P
H
H Hydrophobic
P Hydrophilic or Polar
P
P
P
H
P
P
HH contact energy -1, others 0
K. Lau and K. Dill Theory of protein mutability
and Biogenesis
19
Lattice Bead Model Cont..
0
2
4
1
2
2
1
0
3
3
1
1
2
1
2
0 Left Turn 1 Right Turn 2 Straight
Conformation Vector 02
20
Using Genetic Algorithm

• In this study, the genes represent directions
  taken from one bead on the lattice to another.
• Fitness Min (Energy)
• Crossover
• Mutation

21
CROSSOVER
22
Mutation

• 90 Degree rotation at a bead

23
Mutation

• 180 degrees rotation at a bead

24
Mutation

• Snake (all beads shift one position)

25
Mutation

• Crank Shaft

26
Mutation

• Move back one position

27
Improvements to the model

• Duplicate Predator
• Brood Selection
• Fitness Reducing Roulette Wheel

28
Experiment (Dr. Graham Cox)

• Crossover Rate 1.0
• Population Size 200
• Number of generations 300
• Mutation Rate 0.1

29
Results
Unger Moult
Graham Cox
30
Future Work..

• Extend the model to 3 D.
• Investigate the benefits of parallel genetic
  algorithms
• Explore why is it difficult to get a global
  minima for some sequences (e.g. HP-48)
• Replace the left, right, straight with angles

31
Why Protein Folding?
Simulating 100 microseconds of protein folding
could take 1,025 machine instructions, a
computation that would take 3 years on even a
PetaFLOPS SYSTEM OR keep a 3.2GHz microprocessor
busy for the next million centuries
From Communications of the Bioinformatics,
November 2004 Volume 47, Number 11
32
References

• Development and optimization of a novel genetic
  algorithm for studying model protein folding
  Graham Cox (2004)
• http//faculty.biu.ac.il/unger/jmbga.pdf

33
Questions??