Graphics Application Lab Genomic Sequence alignments and its application 조환규 교수 부산대학교 공과대학 정보 컴퓨터 공학부 [email protected]

Graphics Application Lab

Genomic Sequence alignments and its

application

조환규 교수 부산대학교 공과대학정보 컴퓨터 공학부[email protected]

2

Gra

ph

ics A

pp

licatio

n

Lab

Biology and Informatics

Mathematics : Physics = X : Biology X = ? Bioinformatics

o Understanding Biological System with Informaticso Molecular Biologyo Computational Biology

• Genomics• Proteomics• And several –omics

3

Gra

ph

ics A

pp

licatio

n

Lab

Main features of This Talk

이미 잘 정리된 Computing methodology 를 어떻게 Bioinformatics 에서 활용하는가 ?

Bioinformatics 에서 잘 정리된 방법론은 CS쪽에서 어떻게 활용하는가 ?

Case Study• Genomic sequencing alignment 와 program-copy detection과의 연관

4

Gra

ph

ics A

pp

licatio

n

Lab

Computing Space Transform

Normal Space Jewelry Space

PLUS

MINUS

5

Gra

ph

ics A

pp

licatio

n

Lab

Computing Space Transform(2)

Normal Space log Space

X , Y

X * Y

multiply

log X , log ylog

exponent

log x + log y

6

Gra

ph

ics A

pp

licatio

n

Lab

Computing Space Transform(3)

Program Space Genome Seq Space

Similarity( , )

Similarity( , )

CLUSTAL-W( a, b )

Pairwise Alignment

Protein a Protein bProgram a Program b

Basic keyword

7

Gra

ph

ics A

pp

licatio

n

Lab

Namely “ 우물론” 한 우물을 팔 것인가 ?

o 그러나 만일 끝끝내 물이 안 나올 경우라면 여러 우물을 돌아가며 조금씩 팔 것인가 ?

o 이것도 저것도 아니라면 ? 그렇다면 어떻게 팔 것인가 ?

o Avoid “reinventing wheel”

8

Gra

ph

ics A

pp

licatio

n

Lab

Genomic Sequence Alignments

Genomic Sequences, linearo DNA, RNA, Protein(amino acids)o Why linear ?

Goal of Molecular Biology or Life Scienceo Characterizing functions of geneso Understanding internal gene interactionso Understanding internal & external interactions

• Drug targeting(protein interaction)

Why Alignment ?

9

Gra

ph

ics A

pp

licatio

n

Lab

Human Binome BANK in 3280

Year 3280, dooms day So many binary data files

o Chips( cell-phone, cooker, TV… et al)o Computer disks

Figure out the contents of followingso 010111010101000100101000000001010101010….

?o 100000001010111111111001010101010010101…

?o 101011111111111110001010010100101010100…

?

10

Gra

ph

ics A

pp

licatio

n

Lab

Human-Binome-Project

They decided to establish Binary BANK.o Some are partially annotated

• Object code, text data , garbage• Protein , Gene , Junk-DNA

HUMAN-Binome PROJECT….. Starts…o 목적 : 각종 binary sequence 의 기능을 탐색o Mini-binary project ~~ E.Coli, C.Elegans

• Cell-phone, calculator, PDA…o Full sequencing of 300 Giga bytes DISK

HUMAN BINOME BANK(HBB)!

11

Gra

ph

ics A

pp

licatio

n

Lab

For an Unknown Seq. X

Sequence X from a hardwareo Several error bits includedo Fragment sequencing

Find a similar pattern in HBBo Function ?o Region ?o Size ?

Write a BIG paper….o Here it comes………

12

Gra

ph

ics A

pp

licatio

n

Lab

반도지역 SM-6-5 에서 흔히 발견되는Seq-45-6-X 종의 기능에 대한 고찰

팔봉이 , 2 급 원숭이우주대학 , 분자고생물학과

이번에 SM-6-5 지역에서 빈번하게 발견된 유전종들의 기능은 … . 아마도 이 부족들이 집중적으로 사용한 언어를 표기하기 위한 도구의 일부로 추정된다 . 또한 이들은 매우 다양한 표기형식을 가지고 있는 것으로

추정되는데 , 그 이유는 각 단어들이 나타나는 빈도에서 추정되는 엔트로피가 PR-99 지역에 나타나는 약 26 개보다 월등히 많은 곳으로

추정되므로 …………………… .

Journal of Science & Sciences No.187 (Vol. 213), pp.232-265, 3288

13

Gra

ph

ics A

pp

licatio

n

Lab

Dynamic Programming

A Basic Methodologyo For all kinds of alignment o Solution from all sub-partial solution

준비물o Objective functiono Dynamic programming formula(recursion)

• F(n) = F(n-1) + F(n-2),

o Base condition, F(0)=F(1)=1o Table, multi-dim. array structure

14

Gra

ph

ics A

pp

licatio

n

Lab

Global Alignment(1)

Basic scoring: o Match: 1, Mismatch: -1, Space: -2

How? o To find the alignment of two sequences of maximal sco

re

o Sequence alignment problem corresponds to the longest path problem form the source to the sink in this directed acyclic graph.

15

Gra

ph

ics A

pp

licatio

n

Lab

Global Alignment(2)

CACAGTGT 와 CAGGT

-2

-4

-6

-8

-10

-20 -12-6 -8 -10-4 -14 -16

1

-1

-1

2

-3

0

-5 -2

-3 0

-5 -7 -9

-2 -4 -6

1 -1 -1 -3

-1 0 0 -2

-7 -4

-11 -13

-8 -10

-5 -7

-2 -4

-3 -2 -1 1 -1 -1

C A C A G T G T

C

A

G

G

T

T

T

G

G

A

A

G

G

T

-

A

-

C

-

C

C

16

Gra

ph

ics A

pp

licatio

n

Lab

Local Alignment(1)

An alignment between a substring of s and a substring of to Each entry of (I,j) will hold the highest score of an align

ment between a suffix of s[1…i] and a suffix of t[1…j]

예 ) AGGTATTGA - CCTATGGC

17

Gra

ph

ics A

pp

licatio

n

Lab

Local Alignment(2)

AGGTATTG 와 CTATGC

0

0

0

0

0

00 00 0 00 0 0

0

0

0

0

0

0

0

1

0

0

0 0

1 0

0 0 2

0 1 0

A G G T A T T A

C

T

A

T

G

0

0 1

0 0 0

1 1 0

0 0 2

3 2 0

1 0 0 1 2 1

0 0 0 0 0 0 0 1C

A G G

- - C

T

T

A

A

T T A

T G C

18

Gra

ph

ics A

pp

licatio

n

Lab

Semi-global Alignment(1)

Given two sequences, check if one of them has a substring similar to the other entire sequence.

How? o Find alignments ignoring the beginning and end spaces

of the sequences Global alignment 와 비교

o CAGCA - CTTGGATTCTCGG <-semi-globalo - - -CAGCGTGG- - - - - - - (score: -19)

o CAGCACTTGGATTCTCGG <-global o CAGC- - - - -G -T- - - -GG (score: -12)

19

Gra

ph

ics A

pp

licatio

n

Lab

Semi-global Alignment(2)

CACAGTGT 와 CAGGT

-2

-4

-6

-8

-10

-20 -12-6 -8 -10-4 -14 -16

1

-1

-1

2

-3

0

-7 -4

-11 -13

-8 -10

-5 -7

-2 -4

-3 -2 -1 1 -1 -1

C A C A G T G T

C

A

G

G

T

C

C

A

A

C A

--

G

G

-5

-3

-2

0

-5 -7 -9

-2 -4 -6

1 -1 -1 -3

-1 0 0 -2

G

- T

T

G T

--

20

Gra

ph

ics A

pp

licatio

n

Lab

General Gap Penalty

Definitiono Gap: consecutive number k > 1 of spaces o When mutations are involved, the occurrence of

a gap with k spaces is more probable than the occurrence of k isolated spaces

o w(k) : penalty associated with a gap with k spaces

21

Gra

ph

ics A

pp

licatio

n

Lab

Affine Gap Penalty Function

Penalty for consecutive spaces <= isolated spaces Sub-additive function

o w(k1 +k2+…_kn) <= w(k1) + w(k2) +…+w(kn) Three arrays for dynamic programming

o a[i,j] = maximum score of an alignment between s[1…i] and t[1…i] that ends in s[i] matched with t[j]

o b[i,j] = maximum score of an alignment between s[1…i] and t[1…i] that ends in a space matched with t[j]

o c[i,j] = maximum score of an alignment between s[1…i] and t[1…i] that ends in s[i] matched with a space

22

Gra

ph

ics A

pp

licatio

n

Lab

Heuristic Alignment

Main difficulties o Search space, O(n^2) space or O(n^2 log n)timeo Optimality or Biologically-good Distance metrico Multiple alignment

Local searcho Diagonal region searchingo Visualization., e.g., Dotlet

BLAST approach for long sequenceo Small word matchingo And Extending from a highly matched region

23

Gra

ph

ics A

pp

licatio

n

Lab

Multiple Alignment

o ------AG---T----CGCTGC----o ------AGCGAT--CGCGCTGC---o ---TCGAGGCAA--GCTGCTGC-----o ------GGCGAT----CGCTGC-----

Problem hardness: o Optimal alignment : NP-hard

• Almost kinds of object functionso What if more than 1000 sequence ?

• SPACE COMPELXITY IN PRACTICE Pairwise alignment Star Alignment Tree alignment

24

Gra

ph

ics A

pp

licatio

n

Lab

Why multiple alignment ?

Finding Conserved regions Computer virus phylogeny constructing

• 300 sp./year• More than 10000 sp. : N• Number of files in a system : M > 100000• Detecting a CV takes O( N*M) checks!

tuberculosis• 8 종 , a conserved region, and polymorphic sites• 김철민 교수님 ( 부산의대 ) – 진단용 칩 제작

Phylogeny construct

25

Gra

ph

ics A

pp

licatio

n

Lab

Phylogeny 1 : hard Version

26

Gra

ph

ics A

pp

licatio

n

Lab

Phylogeny 2 : Probable version

27

Gra

ph

ics A

pp

licatio

n

Lab

Constructing Phylogenetic Tree

Distance matrix

Optimal Tree ?• Degree constraint, Steiner points, Quartet

method

A B C D E

A 0 4 17

3 8

B 0 11

5 12

C 0 6 12

D 0 8

E 0

A

B D

E

C

Graphics Application Lab

PART 2: Application

Detecting Source Code Plagiarism

29

Gra

ph

ics A

pp

licatio

n

Lab

Plagiarism, Plagiarism, Plagiarism

Linear Structureo Genomic sequenceso Plain articleso Programso Human behaviors on the time-lineo Time-series data sets

Student Reports Plagiarism Assignment Program copying Where is the original version of this one

? Web searching redundancy elimination

30

Gra

ph

ics A

pp

licatio

n

Lab

지문법 기반 표절 검사 시스템

시스템 검사대상 검사방법 기타Plagiarism.or

g일반문서 비공개 온라인 유료 , 대용량 DB 운영

IntergirGuard 일반문서 비공개 온라인 유료

EVE2 일반문서 비공개 온라인 유료 ,인터넷으로 유사한 문서 검색

CopyCatch 일반문서 문서내의 공통 어휘 빈도수 검사

비온라인

WordCheck 일반문서 문서내의 단어 사용 횟수 검사

비온라인

COPS 일반문서 문장 일치 여부 검사 비온라인 , 문장을 DB 에 저장

SCAM 일반문서 단어 일치 여부 검사 비온라인 , 단어를 DB 에 저장

교수 클럽 일반문서 단어 , 특수문자 , 공식 등 의 빈도수 검사

온라인 유료

SIM 프로그램 소스코드 토큰들의 참조 회수 비교

Siff 프로그램 소스코드 50 개의 대표문자 추출하여 비교

31

Gra

ph

ics A

pp

licatio

n

Lab

구조 기반 표절 검사 시스템

시스템 검사대상 검사방법 기타CHECK Latex 문서 문서의 특성을

tree 로 구성키워드 분포도 ( 지문법 )

을 일부적용

Plague 프로그램 소스코드

Longest Common Subsequence

YAP 프로그램 소스코드

스트링 매칭 방법 사용

YAP3 프로그램 소스코드

Karp-RabinGreedy-String-

Tiling

영어로 된 과제의 표절 검사

MOSS 프로그램 소스코드

스트링 매칭 방법 사용

온라인 시스템

Jplag 프로그램 소스코드

Greedy-String-Tiling

온라인 시스템

32

Gra

ph

ics A

pp

licatio

n

Lab

Fingerprinting Method

Keyword frequency similarityo 특정한 단어의 사용횟수 o Fingerprinting object

• Fixed size fingerprint• Easy to making Database• Quick searching• High false positive rate

Example, fingerprint vector

A c x t u x g r N …..

33

Gra

ph

ics A

pp

licatio

n

Lab

Attacking

Inserting redundant words Shuffling Cons and Pros

o Easy to use in document applicationo Hard to use in program file

Recent trendso Structure-oriented similarity measureo Greedy-Block-Removing methods…o Is this a basic concept of local alignment ?

Sample-Report-Server Building

34

Gra

ph

ics A

pp

licatio

n

Lab

Undergraduate Assignment

Programming Assignment cheating: o 이론적으로는 그 구별방법이 없다 .o Assignment cheating 은 비용이 크다

• Password breaking by Mafia

Assignment 의 출력은 동일하다 .o Correct program 들끼리만 비교

과제에 주어진 시간은 비교적 짧다 (3-4 일 정도 ). 수강생의 수는 적절하다 (300 명 이하 ). 프로그래밍 언어는 모두 동일하다 .

35

Gra

ph

ics A

pp

licatio

n

Lab

Program Cheating Techniques

Complete Copying Variable exchange Garbage code insertion Function transpose Code rewriting(partially) Library code replacing Merging different codes Function resolving Function rewriting

36

Gra

ph

ics A

pp

licatio

n

Lab

Computing Space Transform

Program Space Genome Seq Space

Similarity( , )

Similarity( , )

CLUSTAL-W( a, b )

Pairwise Alignment

Protein a Protein bProgram a Program b

Basic keyword

37

Gra

ph

ics A

pp

licatio

n

Lab

PROGRAM to PROTEIN

Program Languageo Keyword = { int, float, class….. }o Block Structure = “}”, “{“

Program Chromosome o Location independent code, JAVA class, C files

Non-Coding regiono /* this is a sample non-coding region */

Promotero Variable declaration, class definition

DNA = keyword sequence

38

Gra

ph

ics A

pp

licatio

n

Lab

Extracting Program DNA

Syntactic level Semantic level

Program Flow-graph

syntax Real running

Syntactic running

39

Gra

ph

ics A

pp

licatio

n

Lab

Flow-Graph Linearization

A

B

S

R

W

$ A B S R S B W Q %$ A B S R R R R S B W Q %$ A B W W W W Q %$ A W W Q %

Q

40

Gra

ph

ics A

pp

licatio

n

Lab

Example

main( ) {int i, j , k ;…………for( I = 1 . I <= 100 , i++) { ………

if ( ) x = y ; else ……. while( ccccc ) { } x = 23984 ;} // end of for………..

}

int for if = else while =

AGTCGCTTCGAAGCAA

41

Gra

ph

ics A

pp

licatio

n

Lab

Why Protein mapping ?

DNA sequence overlapo if = AA, then = AG, * = GA, return = GGo AAGGA = AG + GA or AA + GG + A

• Ambiguity resolving

20 Amino acid baseso About 20 keywordso 2-3 groups

• polar, non-polar • hydrophobic, hydrophilic • Charged, uncharged• Small, large

42

Gra

ph

ics A

pp

licatio

n

Lab

Amino Acids classification

43

Gra

ph

ics A

pp

licatio

n

Lab

Keyword Mapping Strategy

Convertibility = { for , while } Easy

Convertibility = { for, then } Hard

Convertibility = { if, ‘=‘ } Impossible? Procedure

o Preprocessingo Chromosome arrangemento Keyword selectiono Protein mapping

44

Gra

ph

ics A

pp

licatio

n

Lab

Copy Detecting System

CDS components = [K, M, P, T, A, G ]

o Keyword table 20 keywordo Matching Score matrix borrow from Protein(PAM)o Affine Gap Penalty o Threshold lengtho Alignment Set Scoringo maXimum Gap allowing

45

Gra

ph

ics A

pp

licatio

n

Lab

Experiment Overview

Sample programs “data structure” Students , 60 Programming assignments, 12 set 1 semester On-line evaluation system = ESPA

o Java-based on-line evaluation systemo Due, 1 week

We do not monitor all programs

46

Gra

ph

ics A

pp

licatio

n

Lab

Clustal-W (1)(www2.ebi.ac.uk/clusterw)

Input Fasta file

47

Gra

ph

ics A

pp

licatio

n

Lab

Clustal-W(2) (www2.ebi.ac.uk/clusterw)

Output

48

Gra

ph

ics A

pp

licatio

n

Lab

PhyloDraw (cho et al, Bioinformatics 2001.)

49

Gra

ph

ics A

pp

licatio

n

Lab

Experiment Result 1-0

유사한 그룹이 있는 11 개의 프로그램

50

Gra

ph

ics A

pp

licatio

n

Lab


Unit distance topological Representation

51

Gra

ph

ics A

pp

licatio

n

Lab


Rooted representation

52

Gra

ph

ics A

pp

licatio

n

Lab


Time-dependent dendrogram

53

Gra

ph

ics A

pp

licatio

n

Lab


Time-independent dendrogram

54

Gra

ph

ics A

pp

licatio

n

Lab

Experiment Result 2


55

Gra

ph

ics A

pp

licatio

n

Lab

Experiment Result 3


56

Gra

ph

ics A

pp

licatio

n

Lab

Experiment Result 4


57

Gra

ph

ics A

pp

licatio

n

Lab

Experiment Result 5


58

Gra

ph

ics A

pp

licatio

n

Lab

Experiment Result 6

유사도가 낮은 14 개의 프로그램

59

Gra

ph

ics A

pp

licatio

n

Lab

Another Application

Music Score Plagiarism o Tempo, melody line….o C major, A minor , key-transformation

Credit Card Bankruptcy Alert Drinking Alert

annotated time-line

60

Gra

ph

ics A

pp

licatio

n

Lab

Application

Web searching Engineo Eliminate redundant documentso Eg.) Query = “ 썬베드“ in EMPASS search engine

• 탐색된 상위 10 개의 관련 문서 중에서 8 개는 동일한 문서

o 신문기사 검색에서도 유사한 경우 Original Paper

61

Gra

ph

ics A

pp

licatio

n

Lab

Further Work

Program DNA-Bank server Copying Phylogenetics Building Parametric Method

o Fixed-size fingerprinting = program proteino PAM for program copying behavioro Real Practice

University Report Oracle Music Plagiarism

o Phylogeny tree for Old classical music(Palestria to Brahms) How to linearize a procedure call ?

o Parameter tuningo Procedure call is a sort of directed graph

Fast and moderate size Program DNA

62

Gra

ph

ics A

pp

licatio

n

Lab

Conclusion

Bioinformaticso Bioinformatics

• Bioinformatics Bioinformatics

Bioinformatics A brave new world……

Linear Structure Similarity o Local Alignmento Gap penalty o Structure-based similarity

Good Application

63

Gra

ph

ics A

pp

licatio

n

Lab

PUSAN BIOINFORMATICS JIHAD

64

Gra

ph

ics A

pp

licatio

n

Lab

Realtime Home-Bioinformatics

Documents

Graphics Application Lab Genomic Sequence alignments and its application 조환규 교수 부산대학교 공과대학 정보 컴퓨터 공학부 [email protected]