Transcriptional regulatory networks in

1
Transcriptional regulatory networks in
Saccharomyces cerevisiae
Lee et al (19 co-authors) (Richard Young at
MIT) Science 298 799-804. 2002 18 pages of
suppl. data
First sentence of abstract We have determined
how most of the transcriptional regulators
encoded in the eukaryote Saccharomyces
cerevisiae associate with genes across the
genome in living cells.
Method Genome-wide location analysis
2

• 141 TFs listed in Yeast Proteome database
• All endogenous genes for these TFs were tagged at
  3 end with several copies of the Myc epitope
• Obtained 124/141 viable strains and 106 of the
  124 tagged TFs were expressed when grown in rich
  medium.
• Also have a Yeast Intergenic DNA Array. Using the
  Yeast Intergenic Region Primer set (Research
  Genetics) they PCR amplified and printed 6361
  spots, representing essentially all known
  intergenic regions. The average size of the
  spotted PCR products was 480 bp, and the sizes
  ranged from 60 bp to 1500 bp.

3
Crosslink protein to DNA in vivo with
formaldehyde Break open cells and shear
DNA Immunoprecipitate Reverse-crosslinks, blunt
DNA and ligate to unidirectional
linkers LM-PCR Hybridize to array
Did three experiments for each of the 106 strains
4
Conducted statistical tests, etc. to derive a
confidence value (P value) for each spot using
error models
5
At P value 0.001, 2343/6270 (37) genes were
bound by one or more TF. The false positive rate
was experimentally estimated to be 6-10.
Conversely, it was estimated that 2300 true
interactions are missed by using such a stringent
P value. .
6
At P 0.001
data
random distribution
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
Autoregulation
Multi-component loop
Feedforward loop
10 TFs
3 TFs
39 TFs in 49 loops involving 240 genes
12
Multi-Input motif
Single input motif
295 combinations of 2 or more TFs
13
Regulator chain
188 chain motifs each with 3-10 TFs
14
What about larger regulatory networks

• Developed algorithm that combines genome-wide
  location data with expression data to identify
  groups of genes that are both coordinately bound
  and coordinately expressed
• Assaying the combined probability of a set of TFs
  being bound and relying on similarity of
  expression patterns allows the P values of
  individual binding events to be relaxed - which
  recaptures info lost because of an arbitrary P
  value threshold.
• Resulting sets of TFs and genes are multi-input
  motifs refined for common expression (MIM-CE)

15
Regulatory network structure of the cell
cycle MIM-CEs enriched for genes whose exp.
oscillates through the cell cycle 11 TFs
Generated new set of MIM-CEs using the 11 TFs
and cell cycle exp. data Aligned MIM-CEs
around cell cycle on basis of peak exp. of genes
in the group This correctly assigned all TFs to
cell cycle stages where they are known to
function
16
(No Transcript)
17
(No Transcript)
18

• Current dataset - only 2343/6270 genes (37) were
  bound by a TF at p0.001
• presence of epitope tag could alter binding
  properties.
• Microarray defective?
• Found 5 of 6600 spots did not bind genomic DNA
• Too many false negatives.
• Estimates of false negative rates suggest 69 of
  genes may be bound by at least one TF
• Extracellular environment.
• 35 TFs were assayed in one other growth
  condition and 478 additional genes now bound by
  a TF