Affymetrix GeneChip Data Analysis

1
Affymetrix GeneChip Data Analysis

• Chip concepts and array design
• Improving intensity estimation from probe pairs
  level
• Clustering
• Motif discovering and preliminary result

2
How is Affymetrix Gene Chip made?
3
Affymetrix Gene Chip Procedure

• Target preparation
• Isolate mRNA ?cDNA ? cRNA? 35200 base
  fragments
• Target Hybridization
• 200ml cocktail ? probe array chip ?
  hybridization
• (16 hours at 4060C)
• Washing and Staining
• Probe array scan, Data analysis, and
  Interpretation

(12-15ug)
biotinylated
4
Concepts of Array Design

• PM to maximize hybridization
• MM to ascertain the degree of
  cross-hybridization

5
Signal Intensity Estimation

• Absolute analysis
• average difference ?(PM-MM) / pairs
  used
• Comparison analysis
• experimental subtract baseline average
  difference
• Probe pairs variation
• probe pairs outlier (cross-hybridization)
• arrays outlier (image contamination)
• Probe pairs level adjustment
• Wing Wong Lab dChip least square fitting model
  built from multiple chips
• Affymetrix MAS 5.0 imputation

6
Clustering

• Affymetrix DMT
• examine the SOM, are the clusters tight ?
• experiment primarily with "Number of Epochs",
• "Alpha_i" and "Sigma_i" parameters to reduce
  error.
• multiple level clustering
• examine filtering parameters to adjust min
  and max value.
• (default is 20 and 20000)
• Wing H. Wong Lab dChip
• from image file to outlier detection and
  clustering
• MIT Whitehead Institute GeneCluster
• Stanford Cluster and Treeview
• two way clustering (genes and/or arrays)
• find profile pattern not found in SOM clustering

7
Motif discovering

• Motif discovery and matching tools
• Basic assumption a cluster of co-regulated
  genes is regulated by the same transcription
  factors and the genes of a given cluster share
  common regulatory motifs
• Gibbs Motif Sampler
• Neuwald et al. 1995
• AlignACE
• based on Gibbs Sampling, optimized for
  finding multiple motifs and for both strands
• RSA-tools
• motif finding among non-coding regions in
  cress, yeast, fly, worm
• PlantCARE
• search known motifs among 417 different plant
  transcription sites
• TFSEARCH
• search known motifs in vertebrate,
  arthropod, plant, yeast

8
Motif discovering result

• AlignACE motif finding result for a specific
  clustered genes
• 1.3e01 7.9e-04 3.2e-03 153 GGT-TGGWT
• 8.6e00 5.5e-01 4.4e-03 27 GR---AGAA---G
• 7.0e00 5.5e-02 1.2e-03 159 G-G-R--G--GRA
• 6.5e00 1.1e-02 2.2e-02 460 AGATCCA
• 5.7e00 6.8e-01 7.2e-03 538 C-AA-C-AAA
• 1.9e00 9.0e-02 1.5e-02 174 CWTTGGG
• Functional category
• refer to MIPS or TIGR or Gene Ontology web site
• Exhaustive search result
• fetch all possible patterns within a cluster
• compare motif frequency among
• random sampling and functional group
• (applying M. Eisens Cluster/Treeview)

9
Supportive Information

• Schroeder lab (UCSD)
• sequence retrieval and direct MEME motif finding
• Gene Microarray Pathway Profiler
• genechip array statistical consideration
• MIT CCR HHMI Biopolymers Laboratory  
• Databases and software tools for gene expression
• software, databases, mail lists