Statistical Methods in Bioinformatics Biology as Information Science ...

Biostat 226 (TICR Biostat V) Winter 2007, Special Topic II
Statistical Methods in Bioinformatics
Ru-Fang Yeh, PhD, UCSF Biostatistics
Biology as Information Science
• Lecture II.1: An Introduction to Bioinformatics
- Genomics & bioinformatics in clinical research
- Introduction to high-throughput technology & statistical issues
• Lecture II.2: Statistical methods for microarray data
analysis
- Finding association between expression and phenotypes
- Multiple testing
- Sample size calculation
• Lecture II.3: Statistical data mining
- Clustering, classification & prediction
sequence
expression
structure
function
system
4 bases
A:T C:G
4 bases
A:U C:G
20 amino acids
~10000 domains
~300 folds
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Biology as Information Science: in Large Scale!
sequence
Genome
DNA
Human Genome Project:
define parts list, a periodic table of human biology
RNA
expression
Transcriptome
Protein
structure
function
Proteome
system
Systems biology
(Regulome, Epigenome, Metabolome,Physiome...)
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
1

A brief history of the Human Genome Project
How to sequence a genome
• 1859 Darwin’s “The origin of species” - natural selection & evolution
• 1865 Mendel’s peas – “genes”
• 1910 Morgan’s flies – “chromosomes”
• 1944 genes made of DNA
• 1953 Watson & Crick – DNA double helix with base pairing
• 1972 Berg - first recombinant DNA
• 1977 Sanger, Maxam & Gilbert – DNA sequencing methods
First genome (bacteriaphage virus) sequenced
• 1985 Mullis – PCR
• 1986 Hood – automated sequencer
Idea of a large genome sequencing effort
• 1987 Olson - YAC
First genetic map published
• 1988 Human Genome Project initiated: $3 billion, 15 year plan 1990-2005
• 1990 HGP officially began
Venter – shotgun sequencing a gene
Lipman et al - BLAST
• 1992 First physical map of chromosomes
• 1995 H. influenzae genome sequenced
• 1997 S.cerevisiae genome sequenced
• 1998 C. elegans genome sequenced
- Celera founded and entered the sequencing race
Chromosome 22 completed
• 1999
• 2000 D.melagaster genome sequenced jointly by Celera & BDGP
• 2001 “Draft” human genome published; A. thaliana genome completed
• 2002 Draft mouse genome published; Plasmodium & mosquito genome
• 2003 Completion of the human genome
Whole-genome shotgun
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Theoretical progress in shotgun sequencing
Technology Advances
Moore’s law
Require 8-10x sequence redundancy for coverage > 99.9%
From Shendure et al. 2004. Nat Rev Genet 5: 335-344.
Advanced sequencing technologies: methods and goals.
HGP began
~ $1/bp
HGP completed
~ $0.01/bp
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
2

The Last Decade of Progress: Genome Sequencing
$3 billions 1995 H. influenzae 1.8 Mb
What’s Next?
1996
S. cerevisiae 12 Mb
1997
E. coli 5 Mb
1998
C. elegans 100 Mb
Completed human
chromosome
Draft model
organisms
1999
22
2000
2001
D. melanogaster 122 Mb
A. thaliana 115 Mb
Human genome draft
21
20
2002
2003
Human genome
3000 Mb
14 Y 7 6
Mouse
Fugu rubripes
2004
13 19 10 9 5 16
Rat
$10 millions
2005
X 4 2
Chimpanzee
Dog
From Collins et al. 2003. Nature 422: 835-847.
A vision for the future of genomics research.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
What’s Next?
Implications of Genomics in Clinical Research
• ENCyclopedia Of DNA Elements (ENCODE)
- Functional elements/units of DNA
• The International HapMap Project
- Genetic variation (SNPs) among individuals
• Chemical Genomics
- Libraries of small molecules chemical compounds for screening
• Genomes to Life
- Understand how single-cell microbes function
• Structural Genomics Consortium
- 3-dimensional structures of proteins
• Cancer Genome Anatomy/Atlas Project
- Molecular profiling normal, pre-cancer and cancer cells
• Understand disease processes
- Genetic determinants
- Molecular diagnosis
- Guiding choice of therapy
- Develop molecular target drug
- Develop gene therapy
• Determine genetic predisposition to diseases
- Individualized therapy
- Prevention
- Redefine ethnicity
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
3

Finding Genetic Factors in Complex Diseases
Example: Age-Related Macular Degeneration
ARMD (cont.)
• Leading case of blindness in the developed world
• Characterized by progressive destruction of the retina’s central
region (macula: cone cells), causing central field visual loss
• Complex disease --
Known risk factors: age, smoking, diet (lipid intake), family history
• Linkage: 1q31 and other regions
• 3 independent case-control association studies in non-Hispanic
white population all found strongest association with the
complement factor H gene (CFH), then used haplotype analyses to
identify the specific risk polymorphism Tyr402His. (Klein et al,
Edwards et al, Haines et al, Science 2005.)
• The histidine allele accounts for 20-50% of the overall risk.
Haines JL et al Science 2005
Klein RJ et al Science 2005
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
ARMD (cont.)
Biology makes sense too --
• CFH modulates the complement cascade of immune response, and
the known risk factors (diet, smoking, age) and the clinical hallmarks
(multiple drusen, geographic atrophy, choroidal neovascularization) all
correlate with complement activity.
• The risk allele reduces binding to heparin and C-reactive protein
(CRP), which is know to have elevated serum level in ARMD patients.
Good, now what?
• Evaluate prospective risks for individuals carrying the histidine allele,
and in other ethnic groups.
• Complement activity is responsive to drug therapy and life-style
changes -- as potential treatment/prevention.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Molecular Target Drug
Example I: Gleevec (imatinib mesylate)
• First oncology drug developed with
rational drug design
• Chronic Myeloid Leukemia:
- 1-2 cases per 100,000 per year
- < 20% patients cured with main
treatment options (stem-cell
transplantation)
- 95% of patients have Philadelphia
chromosome t(9,22)(q34;q11) and the
resulting Bcr-Abl fusion protein, which
functions as tyrosine kinase.
• Animal experiments established Bcr-
Abl as cause for CML
• Drug target: tyrosine kinase inhibitor
Eureka!!!
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
4

Imatinib (cont.)
• Imatinib also inhibits other tyrosine kinases:
- the type III transmembrane receptors KIT: gastrointestinal
stromal tumors
- platelet-derived growth factor receptor (PDGFR) β: chronic
myeloproliferative diseases
• Patients with idiopathic hypereosinophilic syndrome
responded to imatinib due to a novel fusion tyrosine
kinase FIP1L1-PDGFRα (Cools et al NEJM 2003)
Molecular Target Drug
Example II: Herceptin (trastuzumab)
• HER-2/neu (erbB-2) gene, the human epidermal growth
factor receptor 2, is amplified in up to 30% of breast
cancers. Overexpression of HER2 protein leads to cell
proliferation and affects tumor progression and the
response of tumors to chemotherapy.
• Herceptin: a monoclonal antibody to HER2 protein.
• FDA-approved for HER2+ metastatic breast tumors; also
showing absolute benefits from the early results of Phase
III trial for HER2+ early stage breast cancer as adjuvant
therapy (Piccart-Gebhart et al NEJM 2005).
• Test HER2 by immunohistochemistry (for protein
expression) or FISH (for gene amplification).
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Reverse vaccinology
Example: serogroup B Neisseria meningitidis (MenB)
Bioterrorism/New Epidemic
Example: Severe Acute Respiratory Symdrome
• SARS outbreak began in Nov 2002 in China’s Guangdong province,
then spread rapidly around the globe through close person-to-person
contact.
• 3/29/2003, Dr. Carlo Urbani, the doctor who first identified SARS, dies
of the illness. By then there were 1400+ cases worldwide, including
50+ deaths.
• In late March, UCSF’s Joe Derisi used his “virus chip”, a microarray
that contains exemplar sequences of ~1000 known viruses, to
determine that a new form of coronavirus was present in the SARS
patient samples.
Fraser CM. 2004. Nat Rev Genet. 5(1):23-33.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
• By mid Apr, the genome of SARS coronavirus was sequenced.
• 2004, SARS vaccine human trials.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
5

Molecular Prognosis
Example: breast cancer
Genome Resource Portals
• NCBI Genbank http://www.ncbi.nlm.nih.gov
• Ensembl http://www.ensembl.org
• UCSC Genome Browser http://www.genome.ucsc.edu
Van de Vijver MJ, He YD et al. NEJM 2002.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Additional Resource
• dbSNP: http://www.ncbi.nlm.nih.gov/SNP/
HapMap: http://hapmap.org/
• OMIM: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=OMIM
Genes and disease:
http://www.ncbi.nlm.nih.gov/books/bv.fcgi?call=bv.View..ShowSection
&rid=gnd.preface.91
• Nature Genome Gateway: http://www.nature.com/genomics/
A User’s Guide to the Human Genome II:
http://www.nature.com/ng/journal/v35/n1s/index.html
Nature Genetics Supplements:
http://www.nature.com/ng/supplements/index.html
UCSF Resources
• Functional Genomics Microarray Cores:
http://arrays.ucsf.edu
http://derisilab.ucsf.edu/core/
http://cancer.ucsf.edu/array/
• Center for Bioinformatics & Molecular Biostatistics:
http://www.biostat.ucsf.edu/cbmb
• Cancer Center Biostatistics Core:
http://cancer.ucsf.edu/biostat
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
6

Reading/References
Course Aims
• Guttmacher AE, Collins FS.
Genomic medicine - a primer.
NEJM 2002 347(1):1512-20.
• Guttmacher AE, Collins FS.
Realizing the promise of genomics in biomedical research.
JAMA. 2005 Sep 21;294(11):1399-402.
Biological Question
Experimental Design
High-throughput Data
Generation
Statistical Question
Statistical Methods
Data Preprocessing
& QC
• Collins FS, Green ED, Guttmacher AE, Guyer MS.
A vision for the future of genomics research.
Nature. 2003 Apr 24;422(6934):835-47.
NO
Data Interpretation
Question answered?
YES
Papers in Nature/JAMA/NEJM…
Data Analysis
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Spotted array
Sequencer
Mass Spectrometry
SAGE
Different
High-throughput
Technology
GeneChip Affymetrix
** * * *
Illumina
Bead Array
Agilent:
Long oligo Ink Jet
Nylon membrane
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
7

Microarray Applications
Statistical Issues
Array
Gene
Expression
Probes
on the array
DNA (cDNA, oligos:
gene representatives)
Targets
to be hybridized
mRNA/cDNA
Large-scale Analysis
of…
transcriptional alterations
Technology dependent:
• Preprocessing & QC
CGH
SNP
Methylation
Promoter
Tiling
Protein
Tissue
DNA (clones, oligos)
DNA (oligos)
DNA (CpG island)
DNA (promoter ~1kb)
DNA
antibody
tissues
DNA
DNA
DNA
(bisulfite-treated)
DNA
(ChIP-enriched)
All of the above
protein
proteins
Genomic changes in cancers
Genotyping; Genomic
changes
Methylation-status in genes
Transcription factor binding
sites; histone modifications
All of the above; sequencing;
gene annotation
Protein expression (ELISA)
Histology; protein expression
(immunohistochemistry)
Biological question dependent:
• Hypothesis testing & Multiplicity
• Classification, prediction
• Clustering
Both technology & biological question dependent:
• Experimental design
• Meta-data integration
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Differential Expression Using Spotted Arrays
Statistical Issues
Example Question:
Identifying σ E -dependent genes in E. coli
(Rhodius et al 2006 PLoS Biol.)
Experiment:
Transcript profiling using spotted arrays in wild type
E. coli K-12 strand (low σ E ) versus over-expressing
σ E strand.
• Experimental design:
- Which array? How many samples? Which to co-hybridize?
• Data preprocessing:
- Quality assessment, image analysis, normalization
• Combining replicates for differential expression:
- Effect estimate, statistical significance, multiple testing
• Annotation:
- Shared characteristics (functional groups, motifs) among DE
genes?
• Data archive:
- Data submission to public database
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
8

Statistical Issues
Experimental Design
Oligo/
• Experimental design:
- Which array? How many samples? Which to co-hybridize?
• Data preprocessing:
- Quality assessment, image analysis, normalization
• Combining replicates for differential expression:
- Effect estimate, statistical significance, multiple testing
• Annotation:
- Shared characteristics (functional groups, motifs) among DE
genes?
• Data archive:
- Data submission to public database
Probe design:
which sequence to
print on the array in
which platform
Target design:
allocation of mRNA
samples to the slides
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Two-Color Spotted Arrays
Experimental Design
Arrayed Library
of cDNA/oligos
PCR amplification
Cell/Tissue Total RNA Labeled cDNA
Printing, Coupling, Denaturing
Hybridization
Laser 1 Laser 2
+ =
Scanning &
Image Analysis
Considerations
• Scientific Aims
• Practical considerations
- Types and amount of mRNA samples
- Number of slides/chips available
• Other information: prior experiments, controls planned,
etc.
• Statistical principle: randomization, replication, local
control.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
9

Two-color array specific design issues
• Direct vs Indirect Comparison?
A
A B
B
average( log(A/B) )
Ref
log(A/Ref) - log(B/Ref)
• Time Course:
T1
Two-color Array Design (cont.)
T2 T3 T4 T5 T6 T7
• K > 2 Conditions
(i) All pairs
A1 A2
(ii) Common reference
A1 A2 A3 A4
Ref
A4 A3
Ref
#arrays needed for the same precision for all pair comparisons
r K(K-1)/2 r K
Reference: Yang YH and Speed TP. 2002. Nat. Rev. Genetics.
Design Issues for cDNA Microarray Experiments.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Preprocessing
Type of “Bias”
• Image Analysis: To identify & extract signal & background
Signal-dependent Spatial Print-tip/Plate
GenePix files
(*.gpr)
Cy5 Fg, Bg
Cy3 Fg, Bg
M
Log 2 (Cy5/Cy3)
• Within-Slide Normalization: To identify and remove
systematic effects not due to real biological signal.
A: Log 2 (Cy5)+log 2 (Cy3) / 2
• Quality Assessment: To identify bad quality arrays/spots.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
10

Use Log (base 2) Intensity for Analysis
Boxplots
1.5 x IQR
1 st quartile
(25%)
Inter-quartile range (IQR)
a robust measure of the variability
median
3 rd quartile
(75%)
outlier
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
MA plot (log Ratio vs log Intensity)
Spatial plots/Heatmaps
Rotate by
45 degree
M (minus) :
Log 2 (Cy5/Cy3)
Ratio
A (add) : Log 2 (Cy5*Cy3) / 2
Intensity
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
11

Normalization
Normalization (Cont.)
Which genes to use
Normalization methods
• All genes on the array.
• Constantly expressed genes.
• Spiked controls (e.g. plant genes).
• Genomic DNA titration series.
• Rank invariant set.
Ratios [two channels], within-slide
-- Median
-- Loess
-- Print-tip / pins
Intensities [separate channel],
within- and between-slide
-- ANOVA
-- Quantile normalization
-- VSN
Global
Median
Loess
(or other smoother)
Assumption: No correlation between M and A, or M and layout.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Normalization (Cont.)
+
Print-tip
+ 2D-loess
Single-channel Normalization
• Necessary for analysis methods that model log-intensities
(eg: Wolfinger et al, Kerr et al)
Two stages:
A) within slide: use two-channel methods to remove spatial bias
b) between all single channels normalization: use quantile
normalization or other Affy normalization methods
+
“Local”
(Wilson et al 2003)
Fitted
Normalized
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
12

Quality Assessment
gpQuality() diagnostic plot
Tools:
• Red/Green overlay images
• Diagnostic plots, especially on sets of positive and
negative controls
• Quantitative statistics: signal to noise ratio,
mean/median, variances…
Software: R/Bioconductor
> library(arrayQuality)
> gpQuality(organism=“Mm”)
Bad:
high bg
Good:
low bg
M vs A
Normalized
M vs A
M
M
A M
A A
Pos/Neg Controls
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Software for Preprocessing
• R / Bioconductor:
marray maNorm
limma normalizeWithinArrays, normalizeBetweenArrays
arrayQuality gpQuality
vsn
• GUI:
limmaGUI http://bioinfo.wehi.edu.au/limmaGUI/
• Most commercial and free gene expression analysis
packages offer standard
- Global normalization
- (Print-tip) loess normalization.
Preprocessing Reference
Bioconductor book:
Bioinformatics and Computational Biology Solutions
Using R and Bioconductor. Edited by R Gentleman,
V Carey et al. Springer-Verlag New York. 2005.
Y. H. Yang, S. Dudoit, P. Luu, D. M. Lin, V. Peng, J. Ngai, and T. P. Speed (2002).
Normalization for cDNA microarray data: a robust composite method addressing
single and multiple slide systematic variation. Nucleic Acids Research, Vol. 30, No.
4, e15.
W Huber, A von Heydebreck, H Sueltmann, A Poustka, M Vingron. (2002). Variance
stabilization applied to microarray data calibration and
to the quantification of differential expression. Bioinformatics 18, Suppl. 1, S96-
S104.
Y.H. Yang and N. Thorne (2003) Normalization for Two-color cDNA Microarray Data.
Science and Statistics: A Festschrift for Terry Speed, D. Goldstein (eds.), IMS
Lecture Notes, Monograph Series, Vol 40, pp. 403-418.
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
13

Alternative DNA Microarray Platform: Affymetrix
High Density Short Oligo Arrays
mRNA reference
For one gene (probe set):
11 probe pairs/gene for Human U133plus2 mRNA reference sequence
5’ 3’
…TCGTCTGTATCACAGACACAAAGTTGACTG…
PM: CAGACATAGTGTCTGTGTTTCAACT
MM: CAGACATAGTGTGTGTGTTTCAACT
Probe sequences
Mask design
In Situ Synthesis by Lithography
Perfect Match
MisMatch
…TCGTCTGTATCACAGACACAAAGTTGACTG…
PM: CAGACATAGTGTCTGTGTTTCAACT
MM: CAGACATAGTGTGTGTGTTTCAACT
PM
MM
Fluorescent probe intensity
Cell/Tissue Total RNA Biotin labeled
cRNA
Hybridization
Scanning &
Image Analysis
1.28cm
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Hybridization
+ Scanning
workable raw data
DAT File
[Image, pixel intensities]
Image analysis
dChip
CEL File
[Probe Cell Intensity]
Pre-processing
MAS
GCOS
CDF
+ [Chip Description File]
RMA
Preprocessing of Affymetrix data
Computing expression measures as a three-step procedure:
• Background subtraction (B)
• Normalization (N) to facilitate between-array comparison
• Summarization of 11-20 probe pair (PM/MM) intensities to
one probe set value (S).
Excel File
Intensity Value
CHP File
Intensity value
Absent / Present call
Text File
RMA Value
Log 2
(Intensity)
Let X be CEL file data from multiple arrays then
Expression values = S(N(B(X)))
Data: a matrix of k genes by n samples
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
14

Affymetrix Data Preprocessing Methds
• Affymetrix: MAS v5.1
Signal = TukeyBiweight{log(PM j - MM j *)}
• Robust multi-array analysis (RMA, Irizarry et al 2003):
use median polish or robust regression fit (IRLS) and
estimate background from PMs
log 2 (PM-BG)* ij = chip i + probe j + ε ij
RMA log expression values
• dChip (Li, Schadt & Wong 2001): Maximum likelihood estimate.
PM ij - MM ij = chip i probe j + ε ij
Between-Array Normalization
• MAS5: N/A. Single-chip method. Global scaling suggested
for probe set data.
• dChip: Use piecewise linear normalization of rankinvariant
sets (estimated non-DE genes) between each
array and a baseline array.
• RMA: Quantile normalization
(B Bolstad et al 2003).
dChip expression values
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Affymatrix array Quality Assessment using
weights from affyPLM
Image Gallery:
http://stat-www.berkeley.edu/users/bolstad/PLMImageGallery/index.html
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Affymetrix Preprocessing Softwares
• RMA:
- R/Bioconductor package & functions
affy rma, justRMA
affyPLM fitPLM, image, boxplot
gcrma justGCRMA
- Standalone GUI:
affylmGUI: http://bioinf.wehi.edu.au/affylmGUI/
RMAExpress: http://rmaexpress.bmbolstad.com/
• dChip: http://www.dchip.org
• Affymetrix own software: MAS5, PLIER
Preprocessing Method Comparison: AffyComp
http://affycomp.biostat.jhsph.edu/
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
15

Affymetrix Preprocessing Reference
Bioconductor book
RMA: Irizarry, R. A., B. Hobbs, et al. (2003). "Exploration, normalization, and
summaries of high density oligonucleotide array probe level data." Biostatistics
4(2): 249-64.
GCRMA: Wu Z, Irizarry RA, Gentleman R, Martinez-Murillo F, Spencer F. (2004). A
Model Based Background Adjustment for Oligonucleotide Expression Arrays In
the Journal of the American Statistical Association. 99, 909–917.
Quantile Normalization: Bolstad, B.M., Irizarry R. A., Astrand, M., and Speed, T.P.
(2003), A Comparison of Normalization Methods for High Density
Oligonucleotide Array Data Based on Bias and Variance. Bioinformatics
19(2):185-193
dChip: Li, C. and W. H. Wong (2001). "Model-based analysis of oligonucleotide
arrays: expression index computation and outlier detection." Proc Natl Acad Sci
U S A 98(1): 31-6.
Method comparison
Irizarry RA, Wu Z, Jaffee HA. Comparison of Affymetrix GeneChip expression
measures.Bioinformatics. 2006 Apr 1;22(7):789-94.
Choe SE, Boutros M, Michelson AM, Church GM, Halfon MS. Preferred analysis methods for
Affymetrix GeneChips revealed by a wholly defined control dataset.Genome Biol.
2005;6(2):R16.
Other Statistical Analysis Questions
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
B-cell lymphoma
Subtype discovery by
Clustering
and Class Validation
(Alizadeh et al Nature 2000)
Finding Groups of
Co-Regulated Genes
with similar time
profile in Yeast Cell
Cycle using SOM
Clustering
(Tamayo et al, PNAS 1999, Data from
Cho et al, Mol. Cell 1998)
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
16

Prognosis
Prediction with
Gene Expression:
Classification &
Validation
(van’t Veer LJ et al, Nature
2002; Further validated in
NEJM 2002).
Transcript (Exon)
Detection Using
Tiling Arrays:
Estimation, Testing,
Meta-data Integration
(Kapranov et al. Genome Res 2005.)
Chromosomal
Aberrations by
array CGH:
Segmentation,
Meta-data Integration
(Selzer RR et al. Genes
Chromosomes Cancer. 2005)
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
Genotyping and
Copy Number Estimation
Using SNP Arrays:
Clustering, classification,
estimation
Mass Spectrometry
Proteomic Profiling
Surface Enhanced Laser Desortption/Ionization
Denoising
BB
AB
AA
M/Z
*
Baseline subtraction
Peak identification
Peak alignment
Data matrix:
n sample
x
p M/Z peaks
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
17

(Light) Reading/References
Project
• Quackenbush J. 2006. NEJM 354(23): 2463-72.
Microarray analysis and tumor classification.
• Nature Genetics Supplement:
The Chipping Forecast II (2002), III (2005).
• Gibson G, Muse SV. 2004.
A primer of genome science.
Sinauer Associates.
• Choose a paper/project that fits your own
(clinical/scientific/technology) interest and has an
analytic component
• Proposal due Feb 26 (Monday before Lecture 3)
• Written report due Mar 19 Monday by 5pm via email:
at least two pages, summarizing what you’ve learned
from the paper/critique of the article
• Oral presentation:
- Mar 6: Manjushree Gautam, Nerissa Ko, Andy Choi, Dan Raz
- Mar 13: Ari Green, Zian Tseng, Beth Cohen
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
TICR Biostat V 2007, Topic II -- Statistical Methods in Bioinformatics
18