Transcript Joining the Dots: Network Analysis of Gene Perturbation Screens

Joining the dots…
Network analysis of
gene perturbation data
Florian Markowetz
markowetzlab.org
How to understand a complex
system?
M. mycoides JCVI-syn1.0
Richard Feynman:
“What I cannot create, I do not
understand.”
Functional Genomics:
“What I cannot break, I do not
understand.”
Breaking the system
Drugs
Small molecules
RNAi
Protein
Stress
mRNA
Knockout
DNA
Somatic aberrations
Today’s lecture
• What information do we get out of gene
perturbations?
– Phenotypes and their ‘richness’
• How do we use this information to infer the
internal architecture of a cell?
– Guilt-by-association
– Nested Effects Models
Phenotype: viability versus cell
death
AWT
B-
Phenotype: organism morphology
Boutros and Ahringer, Nat Rev 2008
Phenotype: cell morphology
After gene
silencing
RNAi control
Boutros and Ahringer, Nat Rev 2008
Phenotype: pathway activity
Receptors
BA-
C-
Phenotype: global gene
expression
B-
A-
C-
All the genes in the genome
…
…
…
ABC-
Transcriptional phenotypes by microarrays
Phenotyping produces partslists
Keith Haring, Untitled,
1986
Urs Wehrli, Tidying Up Art, 2003
A challenge
for computation
and statistics
From phenotypes to clusters
A
B
C
A B
From clusters to mechanisms ??
A B
A
B
B
A
A
B
A
B
A
B
Nested Effects Models
Kinase
TF1
TF1
TF2
Nested effect models:
Guilt-by-assocation:
TF2
TF3
subset relations
similarity
Markowetz et al 2005, 2007
Tresch and Markowetz 2008
Nested Effects Models
INPUT
1. Set of candidate pathway genes
2. High-dimensional phenotypic profile, e.g. microarray
Gene perturbations
OUTPUT Graph explaining the phenotypes
Phenotypic profiles
A
B
C
D
E
F
G
H
Inferred pathway
AB
EF
CD
GH
Effects
Anatomy of the NFB pathway
Step 1
Hits
Weak
Phenotype
?
Roland
Schwarz
+ Meyer lab @
MPI IB Berlin
Strong
NFB
Step 2
Knock-down
Known pathway
members
New RNAi Hits
Compare
expression
phenotypes
by NEMs
Nested Effect Models for
NFB
Roland
Schwarz
Take-home messages
• Phenotyping screens probe a cell’s
reaction to targeted perturbations
• Guilt-by-assocation is a powerful predictor
of gene/protein function …
• … but Guilt-by-assocation has limited
ability to infer mechanisms
• Inferring subset relations by Nested Effects
Models provides hierarchical view of cellular
organisation
PLoS Comput Biol 6(2) 2010
the team
Joining the dots
…
Thank you !
Florian Markowetz
markowetzlab.org