Statistical Tools for Accessing

1
Statistical Tools for Accessing Transcriptional
Regulation Christine Steinhoff Computational
Molecular Biology
2
Aspects of Regulation

• Conserved blocks in noncoding upstream regions
  (regulatory candidates)
• Integrative analysis of different datatypes on
  different biological levels (genomic sequence
  RNA protein ...)
• Pathway modelling
• Regulatory aspects of repetitive elements ...
• Regulatory aspects of epigenetic features
• ...

3
Outline Examples from Our Group

• Theory/Statistics - Normalization Issues
• - GO category overrepresentation
• - Integrative analysis of different data
  types
• - ChIP analysis

2. Implementation issues - R/Bioconductor
3. Application and Databases - CORG database
and upstream region analysis
4
Normalization of (microarray) data Variance
stabilization
5
GO Category Overrepresentation

• subgroup of genes from any kind of biological
  experiment
• any functional group overrepresented?
• problem shortcomings of approaches, strong
  dependencies between the categories
• new statistical approach

6
Integrative analysis of different data types
patients
Genomic localization
7
Integrative analysis of different data types
8
ChIP analysis

• Tiling arrays varying affinies from probe to
  probe
• Unspecific binding establish physical model
  (function of stability of DNA complexes)
• physical model explains high percentage of
  variance in the data

9
ChIP analysis
Probes with larger intensity than predicted are
assumed to carry evidence for TF binding
Score of evidence , Which is a
function of Signal intensity Non specific
binding signal Optical background Standard
deviation of all scores in a bin of predicted
intensities containing unspecific binding
ChIP-chip experiments can be used to identify
mammalian in vivo TF binding sites with high
confidence if the probe-specific behavior is
explicitly taken into account.
10
Results de novo motif discovery
MEME
Sequence set
Motif set
CLOVER sequences chromosome 21 22
Significantly enriched motifs
Comparison with known motifs
Candidate motifs
11
Results de novo motif discovery
P53-DO1 GST control found in 23 of 23
SP1 GST control found in 37 of 37
SP1 INPUTcontrol found in 53 of 53
P53-DO1 INPUT control found in 24 of 24
JASPAR
P53-FL GST control found in 11 of 12
JASPAR
12
What is R/Bioconductor ?

• Free software environment for statistical
  computing and graphics
• UNIX platforms,
• Windows
• MacOS
• http//www.r-project.org/

• Open source and open development software
  project for the analysis and comprehension of
  genomic data.
• started in the Fall of 2001.
• Bioconductor short courses.
• All course materials are available on the WWW
• http//www.bioconductor.org/

13
CORG database

• CORG COmparative Regulatory Genomics
• Conservation of non-coding DNA segments across
  multiple homologous genomic sequences
• Pairwise as well as Multiple alignments based on
  the pairwise ones are available.

• Basis for upstream region exploration
• gene structure
• transcriptional start sites
• comparative information
• transcription factor motif annotation

14
CORG database
15
EuTRACC
Repository -gt Array Express

• Storage
• Management
• Retrieval

• Network generation

• Integrative analysis
• Data mining
• Visualization

• What kind of ChIP chips Design issues? Tiling
  Arrays?
• Development of R package for unified analysis
  normalisation issues/implementation of physical
  model
• Further statistical issues Overrepresentation
• Differential behavior
• Cobehavior
• integrative approaches expressionChIPreg
  ulationcovariates

16
Martin Vingron (Head) Abha Singh
Bais (PhD) Ho-Ryun Chung (Postdoc) Szymon M.
Kielbasa (Postdoc) Holger Klein (PhD) Ho-Joon
Lee (PhD) Thomas Manke (Postdoc) Utz
Pape (PhD) Paz Polak (PhD) Hugues
Richard (Postdoc) Marcel Schulz (PhD) Ewa
Szczurek (PhD) Christine Steinhoff (Postdoc) Toma
sz Zemoitel (Postdoc)
Members of the Regulation Group
Thank you for your attention