Promoter and Module Analysis

1
Promoter and Module Analysis

• Statistics for Systems Biology

2
Transcription Factors

• DNA binding proteins that facilitate or inhibit
  Pol II initiation or elongation
• General transcription factors
• Used widely for many genes under many
  circumstances
• Specific transcription factors
• Used to initiate specific genes under specific
  circumstances
• Distinction may not be so sharp!

3
Transcription Factor Families

• Several structures line up amino acids
• Helix-turn-Helix (Homeodomain)
• Helix-loop-helix
• Zinc Finger
• Mostly dimers
• These families have proliferated because of their
  role in attracting transcription apparatus

4
DNA-Binding Proteins

• All proteins interact weakly with DNA
• Proteins with projecting amino acids interact
  with the DNA major groove
• Hydrogen bonds stabilize position of proteins on
  DNA
• Proteins that line up several amino acid contacts
  bind strongly to specific DNA sequences

5
Transcription Factor Recognition Sites

• Typically 6-10 positions very selective and
  several others show bias
• Often selectivity profile summarized by motif

6
Selectivity of Specific T.F.s

• Most TFs recognize 6-10 bases of DNA
• E. coli longer (8-12 bp) TFs
• All sequences are effective
• Yeast areas around promoters selectively cleared
  of nucleosomes
• 30 x accessibility for those
• Animal cooperative binding of several T.F.s

7
Cofactors

• Frequently the effect of DNA-binding proteins
  depends on co-factors
• E.g. ER sits on the DNA but requires estrogen as
  a co-factor to function
• Myc requires Max as a co-factor to stimulate
  transcription
• If Max is coupled with Mad instead, the genes are
  repressed

8
Assembly of Transcription App.

• Change in physical conformation of DNA leads to
  increased likelihood of spontaneous assembly of
  Pol II
• Getting Pol II further into the gene seems to
  require further steps

9
The TF Family Circus
10
Inferring Regulatory Architecture

• Aim to find which regulators influence gene
  expression
• Concerns
• Contributions of many factors to any one gene
• Approaches
• Decision tree (Computer Science)
• Regression (more statistical)
• DNA sequence motifs can be a surrogate

11
The Israeli Module Approach

• Idea model TF binding as a decision-tree
• Steps
• Cluster gene expression profiles
• Fit best regulator tree to each cluster
• Re-assign genes to clusters
• Iterate until converge

12
Strengths and Weaknesses of Module Approach

• Explicitly models interaction among regulators
• Expression arrays give poor estimates of activity
  of TFs or other regulators
• Some regulators could repress genes
• Discrete predictor model is inefficient

13
Update Estimating TF Activity

• Since TF expression data is unreliable for
  activity, could we do better inferring TF
  activity?
• Use DNA sequence motifs as surrogate for TF
  binding
• Fit double E-M complicated!

14
The Regression Approach

• Direct data on TF occupancy from ChIP
• Two stages
• Find candidate TFs by correlation between
  occupancy and sets of genes
• Estimate TF activity in each condition by
  regression model

15
Regression Steps
Preliminary Screen
r gt rthreshold