Post on 11-Jan-2016
transcript
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
1
CS/CBB 545 - Data MiningIntroduction to Bioinformatics
Mark Gerstein, Yale Universitygersteinlab.org/courses/545
(class 2007,01.18 14:30-15:45)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
2
Data Mining• Importance of knowing the Data• Best approaches often require detailed domain
knowledge – non anonomymized aol data, netflix challenge
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
3
Bioinformatics represents one of the biggest "open" areas for mining
• Genomics & Astronomy• Finance, marketing, credit-card fraud• Security and Intelligence
• Relation to experimental sciences
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
4
General Intro. & background on bioinformatics
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
5
Bioinformatics
BiologicalData
ComputerCalculations
+
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
6
What is Bioinformatics?
• (Molecular) Bio - informatics• One idea for a definition?
Bioinformatics is conceptualizing biology in terms of molecules (in the sense of physical-chemistry) and then applying “informatics” techniques (derived from disciplines such as applied math, CS, and statistics) to understand and organize the information associated with these molecules, on a large-scale.
• Bioinformatics is a practical discipline with many applications.
Core
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
7
What is the Information?Molecular Biology as an Information Science
• Central Dogmaof Molecular Biology DNA -> RNA -> Protein -> Phenotype -> DNA
• Molecules Sequence, Structure, Function
• Processes Mechanism, Specificity, Regulation
• Central Paradigmfor Bioinformatics
Genomic Sequence Information -> mRNA (level) -> Protein Sequence -> Protein Structure -> Protein Function -> Phenotype
• Large Amounts of Information Standardized Statistical
•Genetic material•Information transfer (mRNA)•Protein synthesis (tRNA/mRNA)•Some catalytic activity
•Most cellular functions are performed or facilitated by proteins.
•Primary biocatalyst
•Cofactor transport/storage
•Mechanical motion/support
•Immune protection
•Control of growth/differentiation
(idea from D Brutlag, Stanford, graphics from S Strobel)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
8
Molecular Biology Information - DNA
• Raw DNA Sequence Coding or Not? Parse into genes?
4 bases: AGCT ~1 K in a gene, ~2 M in genome
~3 Gb Human
atggcaattaaaattggtatcaatggttttggtcgtatcggccgtatcgtattccgtgcagcacaacaccgtgatgacattgaagttgtaggtattaacgacttaatcgacgttgaatacatggcttatatgttgaaatatgattcaactcacggtcgtttcgacggcactgttgaagtgaaagatggtaacttagtggttaatggtaaaactatccgtgtaactgcagaacgtgatccagcaaacttaaactggggtgcaatcggtgttgatatcgctgttgaagcgactggtttattcttaactgatgaaactgctcgtaaacatatcactgcaggcgcaaaaaaagttgtattaactggcccatctaaagatgcaacccctatgttcgttcgtggtgtaaacttcaacgcatacgcaggtcaagatatcgtttctaacgcatcttgtacaacaaactgtttagctcctttagcacgtgttgttcatgaaactttcggtatcaaagatggtttaatgaccactgttcacgcaacgactgcaactcaaaaaactgtggatggtccatcagctaaagactggcgcggcggccgcggtgcatcacaaaacatcattccatcttcaacaggtgcagcgaaagcagtaggtaaagtattacctgcattaaacggtaaattaactggtatggctttccgtgttccaacgccaaacgtatctgttgttgatttaacagttaatcttgaaaaaccagcttcttatgatgcaatcaaacaagcaatcaaagatgcagcggaaggtaaaacgttcaatggcgaattaaaaggcgtattaggttacactgaagatgctgttgtttctactgacttcaacggttgtgctttaacttctgtatttgatgcagacgctggtatcgcattaactgattctttcgttaaattggtatc . . .
. . . caaaaatagggttaatatgaatctcgatctccattttgttcatcgtattcaacaacaagccaaaactcgtacaaatatgaccgcacttcgctataaagaacacggcttgtggcgagatatctcttggaaaaactttcaagagcaactcaatcaactttctcgagcattgcttgctcacaatattgacgtacaagataaaatcgccatttttgcccataatatggaacgttgggttgttcatgaaactttcggtatcaaagatggtttaatgaccactgttcacgcaacgactacaatcgttgacattgcgaccttacaaattcgagcaatcacagtgcctatttacgcaaccaatacagcccagcaagcagaatttatcctaaatcacgccgatgtaaaaattctcttcgtcggcgatcaagagcaatacgatcaaacattggaaattgctcatcattgtccaaaattacaaaaaattgtagcaatgaaatccaccattcaattacaacaagatcctctttcttgcacttgg
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
9
Central Dogma
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
10
Molecular Biology Information: Protein Sequence
• 20 letter alphabet ACDEFGHIKLMNPQRSTVWY but not BJOUXZ
• Strings of ~300 aa in an average protein (in bacteria), ~200 aa in a domain
• ~200 K known protein sequencesd1dhfa_ LNCIVAVSQNMGIGKNGDLPWPPLRNEFRYFQRMTTTSSVEGKQ-NLVIMGKKTWFSI d8dfr__ LNSIVAVCQNMGIGKDGNLPWPPLRNEYKYFQRMTSTSHVEGKQ-NAVIMGKKTWFSI d4dfra_ ISLIAALAVDRVIGMENAMPWN-LPADLAWFKRNTL--------NKPVIMGRHTWESI d3dfr__ TAFLWAQDRDGLIGKDGHLPWH-LPDDLHYFRAQTV--------GKIMVVGRRTYESF d1dhfa_ LNCIVAVSQNMGIGKNGDLPWPPLRNEFRYFQRMTTTSSVEGKQ-NLVIMGKKTWFSId8dfr__ LNSIVAVCQNMGIGKDGNLPWPPLRNEYKYFQRMTSTSHVEGKQ-NAVIMGKKTWFSId4dfra_ ISLIAALAVDRVIGMENAMPW-NLPADLAWFKRNTLD--------KPVIMGRHTWESId3dfr__ TAFLWAQDRNGLIGKDGHLPW-HLPDDLHYFRAQTVG--------KIMVVGRRTYESF
d1dhfa_ VPEKNRPLKGRINLVLSRELKEPPQGAHFLSRSLDDALKLTEQPELANKVDMVWIVGGSSVYKEAMNHPd8dfr__ VPEKNRPLKDRINIVLSRELKEAPKGAHYLSKSLDDALALLDSPELKSKVDMVWIVGGTAVYKAAMEKPd4dfra_ ---G-RPLPGRKNIILS-SQPGTDDRV-TWVKSVDEAIAACGDVP------EIMVIGGGRVYEQFLPKAd3dfr__ ---PKRPLPERTNVVLTHQEDYQAQGA-VVVHDVAAVFAYAKQHLDQ----ELVIAGGAQIFTAFKDDV d1dhfa_ -PEKNRPLKGRINLVLSRELKEPPQGAHFLSRSLDDALKLTEQPELANKVDMVWIVGGSSVYKEAMNHPd8dfr__ -PEKNRPLKDRINIVLSRELKEAPKGAHYLSKSLDDALALLDSPELKSKVDMVWIVGGTAVYKAAMEKPd4dfra_ -G---RPLPGRKNIILSSSQPGTDDRV-TWVKSVDEAIAACGDVPE-----.IMVIGGGRVYEQFLPKAd3dfr__ -P--KRPLPERTNVVLTHQEDYQAQGA-VVVHDVAAVFAYAKQHLD----QELVIAGGAQIFTAFKDDV
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
11
Molecular Biology Information:Macromolecular Structure
• DNA/RNA/Protein Almost all protein
(RNA Adapted From D Soll Web Page, Right Hand Top Protein from M Levitt web page)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
12
Molecular Biology Information: Protein Structure Details
• Statistics on Number of XYZ triplets 200 residues/domain -> 200 CA atoms, separated by 3.8 A Avg. Residue is Leu: 4 backbone atoms + 4 sidechain atoms, 150 cubic A
o => ~1500 xyz triplets (=8x200) per protein domain 10 K known domain, ~300 folds
ATOM 1 C ACE 0 9.401 30.166 60.595 1.00 49.88 1GKY 67ATOM 2 O ACE 0 10.432 30.832 60.722 1.00 50.35 1GKY 68ATOM 3 CH3 ACE 0 8.876 29.767 59.226 1.00 50.04 1GKY 69ATOM 4 N SER 1 8.753 29.755 61.685 1.00 49.13 1GKY 70ATOM 5 CA SER 1 9.242 30.200 62.974 1.00 46.62 1GKY 71ATOM 6 C SER 1 10.453 29.500 63.579 1.00 41.99 1GKY 72ATOM 7 O SER 1 10.593 29.607 64.814 1.00 43.24 1GKY 73ATOM 8 CB SER 1 8.052 30.189 63.974 1.00 53.00 1GKY 74ATOM 9 OG SER 1 7.294 31.409 63.930 1.00 57.79 1GKY 75ATOM 10 N ARG 2 11.360 28.819 62.827 1.00 36.48 1GKY 76ATOM 11 CA ARG 2 12.548 28.316 63.532 1.00 30.20 1GKY 77ATOM 12 C ARG 2 13.502 29.501 63.500 1.00 25.54 1GKY 78
...ATOM 1444 CB LYS 186 13.836 22.263 57.567 1.00 55.06 1GKY1510ATOM 1445 CG LYS 186 12.422 22.452 58.180 1.00 53.45 1GKY1511ATOM 1446 CD LYS 186 11.531 21.198 58.185 1.00 49.88 1GKY1512ATOM 1447 CE LYS 186 11.452 20.402 56.860 1.00 48.15 1GKY1513ATOM 1448 NZ LYS 186 10.735 21.104 55.811 1.00 48.41 1GKY1514ATOM 1449 OXT LYS 186 16.887 23.841 56.647 1.00 62.94 1GKY1515TER 1450 LYS 186 1GKY1516
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
13
Molecular Biology Information:
Whole Genomes• The Revolution Driving Everything
Fleischmann, R. D., Adams, M. D., White, O., Clayton, R. A., Kirkness, E. F.,
Kerlavage, A. R., Bult, C. J., Tomb, J. F., Dougherty, B. A., Merrick, J. M., McKenney, K., Sutton, G., Fitzhugh, W., Fields, C., Gocayne, J. D., Scott, J., Shirley, R., Liu, L. I., Glodek, A., Kelley, J. M., Weidman, J. F., Phillips, C. A., Spriggs, T., Hedblom, E., Cotton, M. D., Utterback, T. R., Hanna, M. C., Nguyen, D. T., Saudek, D. M., Brandon, R. C., Fine, L. D., Fritchman, J. L., Fuhrmann, J. L., Geoghagen, N. S. M., Gnehm, C. L., McDonald, L. A.,
Small, K. V., Fraser, C. M., Smith, H. O. & Venter, J. C. (1995). "Whole-
genome random sequencing and assembly of Haemophilus influenzae rd."
Science 269: 496-512.
(Picture adapted from TIGR website, http://www.tigr.org)
• Integrative Data1995, HI (bacteria): 1.6 Mb & 1600 genes done
1997, yeast: 13 Mb & ~6000 genes for yeast
1998, worm: ~100Mb with 19 K genes
1999: >30 completed genomes!
2003, human: 3 Gb & 100 K genes...
Genome sequence now accumulate so quickly that, in less than a week, a single laboratory can produce more bits of data than Shakespeare managed in a lifetime, although the latter make better reading.
-- G A Pekso, Nature 401: 115-116 (1999)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
14
Genomes highlight
the Finiteness
of the “Parts” in Biology
Bacteria, 1.6 Mb,
~1600 genes [Science 269: 496]
Eukaryote, 13 Mb,
~6K genes [Nature 387: 1]
1995
1997
1998
Animal, ~100 Mb,
~20K genes [Science 282:
1945]
Human, ~3 Gb,
~25K genes
2000?
real thing, Apr ‘00
‘98 spoof
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
15
Gene Expression Datasets: the Transcriptome
Also: SAGE; Samson and Church, Chips; Aebersold, Protein Expression
Young/Lander, Chips,Abs. Exp.
Brown, array, Rel. Exp. over Timecourse
Snyder, Transposons, Protein Exp.
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
16
2nd gen.,Proteome
Chips (Snyder)
The recent advent and subsequent onslaught of microarray data
1st generation,Expression
Arrays (Brown)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
17
Other Whole-Genome
Experiments
Systematic Knockouts
Winzeler, E. A., Shoemaker, D. D., Astromoff, A., Liang, H., Anderson, K., Andre, B., Bangham, R., Benito, R., Boeke, J. D., Bussey, H., Chu, A. M., Connelly, C., Davis, K., Dietrich, F., Dow, S. W., El Bakkoury, M., Foury, F., Friend, S. H., Gentalen, E., Giaever, G., Hegemann, J. H., Jones, T., Laub, M., Liao, H., Davis, R. W. & et al. (1999). Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285, 901-6
2 hybrids, linkage maps
Hua, S. B., Luo, Y., Qiu, M., Chan, E., Zhou, H. & Zhu, L. (1998). Construction of a modular yeast two-hybrid cDNA library from human EST clones for the human genome protein linkage map. Gene 215, 143-52
For yeast: 6000 x 6000 / 2 ~ 18M interactions
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
18
Large-scale characterization of yeast gene phenotype using
molecular barcodes
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
19
Molecular Biology Information:Other Integrative Data
• Information to understand genomes Metabolic Pathways
(glycolysis), traditional biochemistry
Regulatory Networks Whole Organisms
Phylogeny, traditional zoology
Environments, Habitats, ecology
The Literature (MEDLINE)
• The Future....
(Pathway drawing from P Karp’s EcoCyc, Phylogeny from S J Gould, Dinosaur in a Haystack)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
20
What is Bioinformatics?
• (Molecular) Bio - informatics• One idea for a definition?
Bioinformatics is conceptualizing biology in terms of molecules (in the sense of physical-chemistry) and then applying “informatics” techniques (derived from disciplines such as applied math, CS, and statistics) to understand and organize the information associated with these molecules, on a large-scale.
• Bioinformatics is a practical discipline with many applications.
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
21
Large-scale Information:GenBank Growth
Year Base Pairs Sequences1982 680338 6061983 2274029 24271984 3368765 41751985 5204420 57001986 9615371 99781987 15514776 145841988 23800000 205791989 34762585 287911990 49179285 395331991 71947426 556271992 101008486 786081993 157152442 1434921994 217102462 2152731995 384939485 5556941996 651972984 10212111997 1160300687 17658471998 2008761784 28378971999 3841163011 48645702000 8604221980 7077491
GenBank Data
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
22
Large-scale Information:Explonential Growth of Data Matched by Development of Computer Technology
• CPU vs Disk & Net As important as the
increase in computer speed has been, the ability to store large amounts of information on computers is even more crucial
• Driving Force in Bioinformatics
(Internet picture adaptedfrom D Brutlag, Stanford)
Str
uct
ure
s in
PD
B
0500
10001500200025003000350040004500
1980 1985 1990 19950
20
40
60
80
100
120
1401979 1981 1983 1985 1987 1989 1991 1993 1995
CP
U In
stru
ctio
nT
ime
(ns)Num.
Protein DomainStructures
Internet Hosts
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
23
PubMed publications with title “microarray”
0
500
1000
1500
2000
2500
3000
1998 2000 2002 2004
Per YearCumulative
Nu
mb
er o
f P
aper
s
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
24
Features per Slide
F
eatu
res
per
chip
oligo features
transistors
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
25
The Dropping Cost of Sequencing
• Adapted from Technology Review (Sept./Oct. 2006)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
26
Bioinformatics is born!
(courtesy of Finn Drablos)
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
27
What is Bioinformatics?
• (Molecular) Bio - informatics• One idea for a definition?
Bioinformatics is conceptualizing biology in terms of molecules (in the sense of physical-chemistry) and then applying “informatics” techniques (derived from disciplines such as applied math, CS, and statistics) to understand and organize the information associated with these molecules, on a large-scale.
• Bioinformatics is a practical discipline with many applications.
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
28
Organizing Molecular Biology
Information:Redundancy and
Multiplicity• Different Sequences Have the Same
Structure• Organism has many similar genes• Single Gene May Have Multiple Functions• Genes are grouped into Pathways• Genomic Sequence Redundancy due to the
Genetic Code• How do we find the similarities?.....
(idea from D Brutlag, Stanford)
Integrative Genomics - genes structures functions pathways expression levels regulatory systems ….
Core
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
29
Where does "mining" fit in Science?
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
30
Bioinformatics as New Paradigm forScientific Computing
• Physics
Prediction based on physical principles
EX: Exact Determination of Rocket Trajectory
Emphasizes: Supercomputer, CPU
• Biology Classifying information and
discovering unexpected relationships
EX: Gene Expression Network
Emphasizes: networks, “federated” database
Core
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
31
Statistical Analysis
vs. Classical Physics
Bioinformatics, Genomic Surveys
Vs.
Chemical Understanding, Mechanism, Molecular Biology
How Does Prediction Fit into the Definition?
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
32
Differences between Mining in Science vs Other Contexts
• Biology & Chemistry are Experimental Sciences Goal is construct an experiment that illuminates fundamental
mechanism Correlation is a means not a goal In contrast, in social sciences one can't readily uncover mechanism
• Nevertheless genome scale data changed things So much data that it contained high-dimensional non-obvious
"patterns" that could be teased out by mining
• Data mining is best as "Target Selection" to suggest experiments but if one does good experiments one won't need careful statistics
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
33
Practical Mining
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
34
Outline for the bioinformatics part of the class
• Illustrate a concrete problem in bioinformatics Predicting gene function (and phenotype)
o Representing genes in terms of networkso Network analysis and prediction
• Why networks are useful for representing information Mining as representing complex information relative to simple generative
models Scale-free models for biological networks
• Bayesian Approaches to Network and Function Prediction Predicting protein interactions
• Spectral Approaches to Phenotype Prediction Using PCA to analyze gene expression data and classify cancers
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
35
~ Outline
• Sequences Alignment
o non-exact string matching, gaps Multiple Alignment and Consensus Patterns
o How to align more than one sequence and then fuse the result in a consensus representation
o Transitive Comparisons
o HMMs, Profiles, Motifs Scoring schemes and Matching statistics
o How to tell if a given alignment or match is statistically significant Evolutionary Issues
o Rates of mutation and change
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
36
~ Outline
• Sequence / Structure Secondary Structure “Prediction” Tertiary Structure Prediction
o Fold Recognition
o Threading
o Ab initio
• Structures Basic Protein Geometry and Least-Squares Fitting Docking and Drug Design as Surface Matching
o Calculation of Volume and Surface Structural Alignment
o Aligning sequences on the basis of 3D structure.
(
c) M
Ger
stei
n '0
6, g
erst
ein
.info
/tal
ks
37
Basis for the topic selection
• Sequence and Structure are classic bioinformatics They are the most prevalent problems
• More difficult to represent for mining Require specialized techniques not covered (e.g. HMMs) Networks and function prediction most similar to classic mining
problems
• Covered Sequence and Structure in MBB/CS/CBB 452/752 Didn't cover networks for function prediction