Transcript Systems Biology

Phenotype and the Interaction of
Genetic Perturbations
Informatics for System Genetics
Phenotype and the Interaction of
Genetic Perturbations
- Introduction
- Generalized derivation of genetic-interaction networks
- Generation of a yeast invasiveness network
- Local and global interaction patterns
Phenotype and the Interaction of
Genetic Perturbations
Introduction
Network element activities and
phenotype
• microarray/proteomics: expression and physical
interactions of each constituent
• phenotype: a system variable
• biomedicine
Directed Perturbations
• Many systems have deletion
projects/consortia/databases
– yeast, worm, mouse, fly
• Molecular biology methods can target large
numbers of genes
– antisense oligos, including morpholinos
– RNA interference
– inducible promoters
RNAi
Fraser et al. (Nature 408,
p. 35, 2000) targeted 90%
of genes on C. elegans
chromosome I using RNA
interference experiments,
and classified resulting
phenotypes.
Synthetic Genetic Array analysis
• Systematic
construction of
double deletion
mutants
• A mutant is crossed
to an array of ~5000
deletion mutants.
• Observing synthetic
lethal genetic
interactions,
generated a network
of 291 interactions
between 204 genes.
• Tong et al. (Science
294, p. 2364, 2001)
SGA
Genetic-Interaction Databases
Phenotype Ontologies
Open Microscopy Environment
(Sorger Lab)
What’s Needed
-Parallel advances in in concepts and computational methods
- Generalized derivation of genetic-interaction networks
- Quantitative (at least ordered) phenotype data
- Analysis of local and global interaction patterns
Phenotype and the Interaction of
Genetic Perturbations
Generalized derivation of genetic-interaction networks
Genetic Interaction
- Interaction of two genetic perturbations in the determination
of a phenotype
- Observed in the phenotypes of four genotypes:
1) a reference genotype, the “wild type”
2) a perturbed genotype, A
3) a perturbed genotype, B, with a perturbation of a
different gene
4) a doubly perturbed genotype, AB.
- Perturbations may be of any form (null, loss-of-function,
gain-of-function, dominant-negative, etc.).
- Two perturbations can interact in different ways for different
phenotypes or under different environmental conditions.
Example
Hereford-Hartwell 1974
Hereford-Hartwell double mutant
epistasis analysis
Hartwell: Synthetic Defects and
Phenotype Buffering
Example
75 Phenotype Inequalities in 9
(A)Symmetric Interaction Modes
Phenotype and the Interaction of
Genetic Perturbations
Generation of a yeast invasiveness network
Dimorphic Fungal Pathogens
S. cerevisiae
Magnaporthe grisea
Filamentous-Form System Properties
• altered cell-cycle progression
• cell elongation
• unipolar distal budding
• adhesion
• host (substrate) invasion
• altered metabolism
Key Pathways
Large-scale Genetic Perturbation
Transformation of
Knockout Strains with
Multicopy Plasmids
• We transformed 118
homozygous diploid knockout
strains plus a wildtype control
strain with plasmids for
constitutive overexpression of
genes involved in regulation of
filamentous growth.
Rsr1
Bem1
Ras2
Cdc42
Phd1
Ste11 Kss1
Ste12
Tec1
Flo8
Gln3
Msn1
Phenotype Analysis
Wash Assay for agar invasion:
Strains:
Diploid S1278b mutants transformed with
multicopy plasmids
Phenotype:
prewash colony
•Agar invasion
Conditions:
High glucose, low nitrogen
postwash colony
Strain Construction
MATa
X
xxx
MATa
yyy
Mata xxx::HygMX
x
Mata yyy::KanMX
Mata xxx::NatMX
x
Mata yyy::KanMX
Mata
xxx::HygMX
yyy::KanMX
Mata
xxx::NatMX
yyy::KanMX
Mating
a/a
Sporulation
Haploid Selection
P-MFA1::HIS3
Homozygous Double Mutant
xxx yyy
Mate and select for HygR NatR to get diploid xxx yyy
Phenotype Analysis
Strains are pinned onto solid media in a 384-spot format.
Each strain is represented by 4 independent constructions.
4 replicates of each plate are pinned.
Each plate contains 48 spots of a wildtype vector control strain.
Analysis of agar invasion phenotypes of
diploid mutant strains on low-nitrogen media
Incubate 4
days at 30o C
Pin strains onto low-nitrogen media
Wash plate
Scan plate
Prewash image
Scan washed plate
Postwash image
Quantitation of Invasiveness
Agar invasion phenotypes of diploid mutant strains on low-nitrogen media
Prewash image
Postwash image
flo1
Agar invasion can be
visualized in the
composite image.
flo11
dia4
isw1
dfg16
rim9
bud8
hmi1
bud6
tpk2
Non-Invasive
Invasive
Hyper
Phenotype Data Analysis
I.
Calculate ratios of postwash signal/prewash signal

Raw data file from dapple processed to ID spots and subtract background.

Output contains X = prewash signal and Y = postwash signal for each spot.

Calculate the ratio Y/X for each spot.
II.
III.
Normalize data to allow comparison of strains on different plates

Each plate contains 48 wildtype controls.

Calculate the median Y/X ratio for the wildtype vector controls on each plate
= Mn for plate n.

The correction factor for plate n is [median (all M values)/Mn].
Phenotype Error = MAX(MAD, 10%MAD)
Invasiveness
Quantitative Phenotypes
Example: Image Data
Example: Data Analysis
Phenotype Error
Data Subset
Entire Network
Interaction-Mode Distribution
Error Parameter Insensitivity
Effect of error model on distribution of interaction classes
900
800
700
non-interacting
asynthetic
suppression
synthetic
epistatic
conditional
double nonmonotonic
single nonomonotonic
additive
# of interactions
600
500
400
300
200
100
0
0
10
20
30
40
50
60
Percentile MAD
70
80
90
100
Phenotype and the Interaction of
Genetic Perturbations
Local and global interaction patterns
Local Interaction,
with Biological Processes
- Is there “monochromatic” interaction with modules?
Gene
Form
Interaction
Biological Process
-log10P
PBS2
null
additive
signal transduction
2.99
PBS2
null
additive
small GTPase mediated signal
transduction
2.96
STE12
gf
single-nonmonotonic to
protein targeting
2.87
STE11
da
noninteractive
cell cycle
2.73
PHD1
gf
hypostatic to
invasive growth
2.68
PDE2
null
noninteractive
protein amino acid
phosphorylation
2.56
HSL1
null
suppressed by
cell wall organization and
biogenesis
2.52
STE20
gf
single-nonmonotonic to
protein targeting
2.31
EGT2
null
conditioned by
invasive growth
2.30
ISW1
null
suppresses
small GTPase mediated signal
transduction
2.30
CLB1
null
noninteractive
protein metabolism
2.30
STE11
da
suppresses
cell surface receptor linked
signal transduction
2.28
BEM1
gf
conditioned by
nucleobase, nucleoside,
nucleotide and nucleic acid
metabolism
2.25
PBS2
null
additive
RAS protein signal
transduction
2.24
PBS2
null
additive
sporulation
2.24
TEC1
gf
synthetic
intracellular signaling cascade
2.19
Local Interaction, with Biological Processes
As noted for epistasis and synthesis…the results suggest
there are characteristic network mechanisms to be found
underlying the various modes of genetic interaction.
Global Interaction Patterns
- genetic-interaction complexity
- map similarities among perturbations in interaction patterns
Global Interaction Patterns
Mutual Information
P( x) , where x  X , and
I [ A; B] 

aA
and
bB
P( x)  1.

xX
 P ( a, b) 

P(a, b) log 2 
 P(a ) P(b) 
I [ A; B]  I [ B; A]  0 bits.
Global Interaction Patterns
Gene1a
Gene2a
Commonb
Mutual Info.c
STE20(gf)
STE12(gf)
99
1.8
16.3
PBS2(lf)
HOG1(lf)
101
1.2
14.1
CDC42(gf)
BEM1(gf)
99
1.0
9.5
STE20(gf)
CDC42(gf)
100
1.5
9.2
PBS2(lf)
HSL1(lf)
95
1.5
8.9
STE12(gf)
CDC42(gf)
101
1.5
8.0
FLO8(gf)
STE20(gf)
100
1.3
6.7
STE20(gf)
TEC1(gf)
99
0.9
6.6
STE12(gf)
GLN3(gf)
99
1.4
6.3
TEC1(gf)
BEM1(gf)
95
0.7
5.0
SFL1(lf)
HOG1(lf)
75
0.8
4.8
STE12(gf)
BEM1(gf)
97
0.8
4.4
CDC42(gf)
GLN3(gf)
101
1.3
4.3
HOG1(lf)
HSL1(lf)
99
0.9
4.3
CDC42(gf)
PBS2(lf)
86
1.0
3.5
FKH2(lf)
YAP1(lf)
18
2.2
3.5
TEC1(gf)
CDC42(gf)
99
0.8
3.3
ISW1(lf)
YAP1(lf)
17
2.4
3.3
RGS2(lf)
MID2(lf)
15
2.3
3.3
STE20(gf)
GLN3(gf)
98
1.2
3.3
YJL142C(lf)
YAP1(lf)
17
2.1
3.2
EGT2(lf)
RGS2(lf)
16
2.1
3.1
STE12(gf)
TEC1(gf)
98
0.8
3.0
-log10P
A Mutual-Information Network
A Mutual-Information Network
…suggests mutual information reflects similarities in the global
effects of perturbations on molecular information flows.
PhenotypeGenetics
Priorities
1) continuing advances in quantitative phenotype
measurement and ontologies
2) reinforcement or revision of genetic-interaction mode
definitions based on relevance to network
mechanisms
3) extension of all genetic-interaction modes beyond
phenotype ordering to incorporate parameters
derived from phenotype magnitudes
4) comparative genetic-interaction analyses of multiple
alleles (with different effects on function) of individual
genes to learn how different levels of gene activity
impact the network
Network modeling by iterative refinement
Data Acquisition
Phenotypes
Microarrays
Proteomics,…
Analysis Modules
Network
Refinement
Pathway/
Interaction
Databases
Network
Visualization and
Modeling
Phenotype and the Interaction of
Genetic Perturbations
Informatics for System Genetics
Becky Drees
Vesteinn Thorsson
Greg Carter
Alex Rives
Marisa Raymond
Iliana Avila-Campillo
Paul Shannon
Tim Galitski