This preview shows pages 1–3. Sign up to view the full content.
1
Network inference from dynamic (state)
information
Input:
components; states of components (in time)
Hypotheses:
regulatory framework
Output:
proposed regulatory network
Validation:
capture known interactions
For inference of gene regulatory networks, the most frequently used state
information comes from gene expression arrays (microarrays)
There are several microarray types and methods, for our purposes it
suffices to say that a microarray provides a readout of the relative or
(semi)absolute expression level of each gene in the array.
Inference methods
• Need expression snapshots:
– Clustering analysis
– Bayesian networks
• Need expression timecourse:
Need expression timecourse:
– Continuous – Differential equations
– Discrete  Boolean
• Need other types of information:
– Data mining
Clustering analysis
•
Pairwise correlation of expression levels of two genes across time
or conditions, e.g. by Pearson correlation or Euclidean distance
•
These correlations are then clustered
– Hierarchical
clustering  forms a dendogram
• Successive clusters are formed by aggregation of existing
clusters.
• Difficult to decide cutoff of similarity
– Kclustering
• K  predetermined number of groups
• Decide criteria to group the genes – e.g. Group together
genes with similar correlation coefficient.
• Often used as an exploratory data analysis tool
• More computationally intensive than hierarchical clustering
but optimization can be performed.
Drawback: No insight into the causal relationship
Example of clustering analysis
•
Data set – yeast cell cycle expression, expression ratios
•S
c
o
r
e
m
a
t
r
i
x
M
i,j
– matrix of similarities between expression ratio of each
pair of genes.
•
Two
aggregate matrices
Time delay
Qian
et al
. (2001) J Mol. Bio, 314, 10531066
Inverted
The central idea is to find the maximal
aggregated score
Max(E) offdiagonal – time
shifted relationship
Max(D) diagonal – inverted
relationship
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document 2
Network is constructed using clustering methods
Time delayed
expression
Simultaneous
Qian
et al
. (2001) JMB, 314, 10531066
expression
Inverted relationships
in the expression profiles.
Oncogenic signaling network
• Constructed weighted gene
coexpression network based on
pairwise Pearson correlations
• Hierarchical clustering to
detect groups or modules
of coexpressed genes.
This is the end of the preview. Sign up
to
access the rest of the document.
This document was uploaded on 03/01/2011.
 Fall '09
 Work

Click to edit the document details