Similarities and differences in genomewide expression programs of six organisms Sven Bergmann

1
Similarities and differences in genome-wide
expression programs of six organismsSven Bergmann
2
Large-scale expression data
E.coli
S. cerevisiae
A. thaliana
H. sapiens
C.elegans
D. melanogaster
3
How to extract biological insight from these data?
4
How to integrate sequence information?
5
Comparative analysis

• Practical tools
• decomposing data into transcription modules
• integration of sequence information provides
  hypotheses
• Global analysis
• study global network properties
• reveal design principles

6
Signature Algorithm
7
Signature Algorithm Output
Co-regulated genes
8
Iterative Signature Algorithm
OUTPUT
9
Identification of transcription modules using
many random seeds
10
Threshold parameter allows for modular
decomposition at different resolutions
11
Representation as Module trees
Cell-cycle
Amino-acid metabolism
TM
Stress
Mating
Protein synthesis
12
Color-code visualizes module properties
13
Segregation between modules with and without
homologues
Number of modules
Random control
data
Fraction of homologues
14
Do the six organisms share the same basic
architecture structure?
Distinct modular organization may reflect
different adaptation requirements
15
Comparative analysis

• Practical tools
• decomposing data into transcription modules
• integration of sequence information provides
  hypotheses
• Global analysis
• study global network properties
• reveal design principles

16
Sequence similarity allows for gene mapping
But which homologue has similar function?
17
Mapping Transcription Modules
18
Average correlation Ribosomal genes

• highly correlated
• statistically significant

19
Average correlation
Mapped transcription modules exhibit significant
correlations!
20
Gene Refinement
BLAST
21
Refined modules contain only co-regulated genes
22
Available annotation indicates that added genes
are functionally related
Worm heat-shock
23
Higher order correlations
correlated
anti-correlated
24
Correlation patterns are distinct for each
organism
25
Comparative analysis

• Practical tools
• decomposing data into transcription modules
• integration of sequence information provides
  hypotheses
• Global analysis
• study global network properties
• reveal design principles

26
Constructing an Expression network
27
Constructing an Expression network
28
Connectivity
k number of edges per node
29
Connectivity distribution
30
Connectivity distribution for yeast expression
network
31
Connectivity Distributions for Expression Networks

• highly connected yeast genes
• related to protein synthesis
• rRNA processing

32
Other networks with power-law distributions
actors
WWW
power-grid
n(k)
k
k
k
33
Comparative analysis

• Practical tools
• decomposing data into transcription modules
• integration of sequence information provides
  hypotheses
• Global analysis
• study global network properties
• reveal design principles

34
Theoretical Approaches

• Dynamically evolving networks with preferential
  attachment (rich get richer) Barabasi
  Albert 1999
• Systems with Highly Optimized Tolerance
  (HOT)Carlson Doyle 2000

35
Centrality Homology
Fraction of yeast genes with human homologue
connectivity
36
Centrality Homology
37
Centrality Lethality
38
Clustering-coefficient
Friendships between friends
C
Possible friendships
How many of your friends are also friends amongst
each other?
6/15
39
From global to modular Where do the stripes come
from?
Transcription Modules (TM) (co-regulated sets
of genes) appear as stripes!
C
k
40
C against k diagrams for all expression networks
41
Take-home Messages

• Accumulation of large scale expression data makes
  comparative analysis of transcriptomes possible
• Practical tools
• decomposing data into transcription modules
• gene-mapping provides hypotheses
• Expression networks of six organisms shareglobal
  properties (power-law clustered)