From High Throughput Pull-Downs To Protein Complexes: Building a Model of the Physical Interactome of Yeast Shoshana Wodak Hospital for Sick Children [email protected]Depts. Biochemistry & Medical Genetics and Microbiology University of Toronto Swiss-Prot Fortaleza 2006
26
Embed
From High Throughput Pull-Downs To Protein Complexes: Building a Model of the Physical Interactome of Yeast Shoshana Wodak Hospital for Sick Children [email protected].
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
From High Throughput Pull-DownsTo Protein Complexes:
Building a Model of the Physical Interactome of Yeast
Essential role! Our knowledge about them is limited… -can be rather dynamic entities, with variable life times-their formation is likely regulated at various levels, transcriptional level, post transcriptional modification, degradation…
Our knowledge about them is limited… -can be rather dynamic entities, with variable life times-their formation is likely regulated at various levels, transcriptional level, post transcriptional modification, degradation…
1st step: mapping the physical interactome 1st step: mapping the physical interactome
Krogan et al (2006) -combining data from ≠purifications ≠ different MS techniques -only bait-prey associations -complex ‘training’ procedure -ignored ribosomal proteins(baits)
Collin et al (2006) [Consolidated network] -combined data from Gavin and Krogan -bait-prey + prey-prey associations -new Protein Enrichment (PE) score: augmented version of Gavin’s scores + ‘training’ -> Confidence scores
Gavin et al (2006) -combining data from ≠purifications -bait-prey + prey-prey associations -unbiased statistical procedure, log-odds based
0
500
1000
1500
2000
2500
3000
3500
4000
0 20 40 60 80 100 120 140 160 180 200
# False positive PP interactions
# T
rue
po
siti
ve P
P in
tera
ctio
ns
ConsolidatedGavinKroganMIPS_small_scale
Core data
S = 0.38
MIPS small scale
0
2000
4000
6000
0 1000 2000
Comparing the PPI networks Comparing the PPI networks
2708 proteins7123 interactions
1622 proteins9074 interactions
(I)
(II)
(V)
(III)
(IV)
Validation and Analysis
MALDI/MS LC/MS
Deriving the PPI Network
Identifying Functional Modules
Protein complexes are expected to ‘share’ componentsProtein complexes are expected to ‘share’ components
This information is however currently not available from the purification data. The pulled down complexes representtemporal and spatial averages of the in-vivo distribution.
Markov Cluster Algorithm (MCL)
Enright et al. (2002); Van Dongen S. (2002)
Hierarchic Clustering
by near neighbor contactscore, or neighbor pattern
Simulates random walks within graphsby computing highermoments of contactMatrix =Measures similarityin path lengths 1,2,3,4between nodes inthe graph
Parsing the PPI networkinto densely connectedregions
Common approach:
0
5
10
15
20
25
30
35
40
45
MIPS Consolidated MCL+overlap
Gavin (all) Gavin(core+module)
Mea
n
Overlaps per complexShared genes per overlappingcomplex
Fraction of complexes sharing subunits with other complexes
41.4% 19.5% 96.9% 84.1%
Degree of overlap between complexes computed usingdifferent PPI networks and different methods Degree of overlap between complexes computed usingdifferent PPI networks and different methods
(I)
(II)
(V)
(III)
(IV)
Validation and Analysis
MALDI/MS LC/MS
Deriving the PPI Network
Identifying Functional Modules
87
341
5310
291177
4732
77
71
35
20
209
99
50
42
Overlap ≤ 5%
5% < overlap ≤ 50%
50% < overlap ≤ 90%
90% < overlap
Gavin(491)
Krogan (547)
Consolidated_MCL (400)
Gavin_MCL (203)
00.10.20.30.40.50.60.70.80.9
1
Precision Homogeneity
Score
Gavin
Gavin_MCL
Krogan
This study
0
0.2
0.4
0.6
0.8
1
Gavin
Gavin_MCL
KroganThis study
MIPS
PPV score
024681012
Semantic Similarity (SS)
Co-localization
SS per cluster
Conso
lidat
ed
M
CL
Overlap with MIPS complexes
Overlap with MIPS complexes
Cellular localizationGo annotations
0
0.2
0.4
0.6
0.8
1
Precision Homogeneity
Score
Gavin/Krogan
Gavin_MCL/Krogan
Gavin_PE/Krogan_PE
25
399
4720
13
9858
34 30
140
40
111Overlap ≤ 5%
5% < overlap ≤ 50%
50% < overlap ≤ 90%
90% < overlap
Gavin/Krogan Gavin_MCL/Krogan Gavin_PE/Krogan_PE
(a)
(b)
(491) (547) (203) (547) (321) (640)
Ribosomal Small Subunit
Ribosomal Large Subunit
RNA Pol. I, II, III
19/22S Regulator
20S Proteasome
RSC
Mediator
Exosome
Mitochondrial Ribosome
TFIIIC
MRP RNase
APC
COP I
Golgi Transport
Exocyst
SRP
SNF1
H+ Transporting ATPase, Vacuolar
SAGA
bc
da
GeneProVlasblom et al. (2006)
POL II
POL IIIPOL I
SAGA-like complexTFIID
SAGA complex
ADA complex
Fig. 8c
(I)
(II)
(V)
(III)
(IV)
Validation and Analysis
MALDI/MS LC/MS
Deriving the PPI Network
Identifying Functional Modules
Protein 3D structure
Diffraction Pattern
Phasecalculation
Model refinement
AcknowledgementsAcknowledgements
Shuye Pu (HSC, Toronto)James Vlasblom (HSC, Toronto)Chris Orsi (HSC, Toronto)Mark Superina (HSC, Toronto)Gina Liu (HSC, Toronto)CCB Systems Support team (HSC, Toronto)
Nicolas Simonis (ULB Belgium)Jacques van Helden (ULB, Belgium)Sylvain Brohée (ULB, Belgium)