Home
Inferred
energy
models
Evolutionary
analysis
Datasets and
annotations
Gene
expression correlations
The
PREGO program
|
Extensive
low-affinity TF-DNA interactions in yeast
|
|
|
This website supports the
paper "Extensive low affinity TF-DNA interactions in the yeast genome".
Use it to explore the quantitative binding preferences of specific TFs
and their evolutionary behaviors. The site also contains information on
corrected ChIP annotations and low level normalizations.
|
|
This work shows that noisy-probabilistic
TF-DNA
interactions are suprisingly abundant in-vivo.
Here is the key observation that initiated this study:
- ChIP on chip experiments quantify the strength of the
binding interaction + a lot of experimental noise
- TF-DNA interaction can be predicted from the
promoters sequences (using PWMs) - again with noise
- Assume the TF have just two groups of targets:
deterministic hits
and deterministic "non-hits", then the ChIP to sequence correlation
should look like a
mixture of two populations. Inside each population we should not
observe ChIP-sequence correlation.
- But the actual correlation of ChIP and sequence
is strikingly
different - in many of the cases the correlation affect more than 20%
of the
genes - but only few percents of the genes have deterministic binding
site
for the typical TF.
- The plot to the right shows Gcn4 ChIP (X axis)
and Sequence (Y axis) affinities. The best fit of the data to two
population requires 20% of the genes to be recognized as Gcn4 targets
and suggest strong quantitative correlation between ChIP and Sequence
inside this 20%.
- We can therefore use ChIP data to study weak
transcriptional interactions. The experiments are very noisy - but
contain much more
information then just a list of hits...
|
|
|
|
For comments and
requests, please contact Amos Tanay, Rockefeller University
|
|
|
