1% RRS PRS RRS YI2 YI2 1 Screen YI2 2-3 Screens Lit-BM-13
Alice Desbuleux
1,2, Tiziana Cafarelli
1, Yang Wang
1, Thomas Rolland
1, Noor Jailkhani
1, Yun Shen
1, Katja Luck
1, Kerstin Spirohn
1, Sudharshan Rangarajan
1, Charles Ogagan Jr.
1,
Jean-Claude Twizere
2, Mike Calderwood
1, Tong Hao
1, Yves Jacob
1,3, David Hill
1, Quan Zhong
1,4, Marc Vidal
1, Benoit Charloteaux
11DFCI Center for Cancer Systems Biology (CCSB) and Department of Cancer Biology, and Harvard Medical School, Department of Genetics, Boston, MA, 2University of Liege, GIGA-R, Liege, Belgium ,
3Pasteur Institute, Department of Genetics, Paris, France, 4 Wright State University, Dayton, OH
High-quality ORFeome (HIP collection) 5,143 yeast genes High-quality ORFeome (HIP collection) 5,143 yeast genes
Pairwise test
FiPP 11,286 positive pairs with confirmed ORF identity
Validation in an orthogonal assay : GPCA
YI2
S. cerevisiae
~6k ~6k
Search Space
Alternative Y2H assay : Low copy number plasmid (Cen) vs high copy number plasmid (2μ) Y2H assay DB-1 AD-2 DB DB 1 1 AD 2 2 AD Reporter Comparison of growth to
identify autoactivators Positive colonies picked into selective media
-L-W
3AT -LWH 3AT CHX -LH
autoactivators positive colonies SWIM-seq
First pass screen (x3)
X
...
3,766 first pass pairs (FiPPs) identified after 3 screens
AD-700 DB-1 AD-2 AD-2 AD-3 AD-4 AD-3
Mate 1 DB against libraries of ~700 AD
DB-1
...
AD-2 X FiPP 3,766 DB-1 X AD-2 1 2 3 4 5 6 7 0 10 20 30 40Score threshold (arbitrary units)
Fr
action of pairs rec
ov
ered (%)
~6k ~6k
Complete search space
YI2
~6k
~6k YI2 search space
Proto-genes
Binary PPI maps
0.05 0 0.1 0.15 0.2 0.25 0.3 0.35 PRS RRS Lit-BM-13 YI1 Ito-Core Uetz Tarrassov Y2H GPCA
Recovery rate of various binary PPI datasets
present in Lit-BM-13 tested in Y2H positive in Y2H 0 5 10 15 20 25 30 YPL122C YIL143C YER171W YGR258C YLR005W YPR056W YDR311W YOR244W YEL018W YJL081C YNL136W YBR288C YGR261C YNL230C YJR052W YJL036W YDL113C YGR163W YBR077C YLR192C YMR012W YMR146C YBR079C YOR361C YMR309C YDR429C YGR020C YKL080W YOR270C YDL185W YEL051W YPR036W YLR447C YOR332W YHR026W YPL234C YHR039C-A YEL027W YBR127C YPR023C YGR002C YNL107W YHR090C YDR359C YJR082C YFL018C YNL071W YGR193C YER178W YBR221C YIL042C YGR003W YBR114W YPL046C YKL041W YMR077C YLR025W YPL195W H+-transporting ATPase, vacuolar NuA4 histone acetyltransferase complex eIF3 nucleotide-excision repair factor 3 complex
mitochondrial pyruvate
dehydrogenase complex Cul3p-RING ubiquitinligase complex
AP-3 adaptor complex ESCRT III complex EGO complex Atg20p/Snx4p/Snx41p complex
1screen 2screens 3screens RRS PRS
AA
Negative Positive
Pairs from screens
Nb Pairs 0 500 1000 1500 2000 2500 3000 0 1 2 3 0 1000 2000 3000 4000 Number of screens Number of pairs
First pass pairs
Y2H verified pairs
0 5 10 15 20 25 30 Small-scale Systematic
Assembly from databases
(min 2 supporting pieces of evidence) Lit-BM-13
Uetz et al., Nature 2000 (Uetz) Ito et al., PNAS 2001 (Ito Core) Yu et al., Science 2008 (YI1) Tarrassov et al., Science 2008
Co-complex datasets Pu et al., Nucleic Acids Res 2008 (CYC2008): catalogue of 408 heteromeric protein complexes manually curated from the literature
Collins et al., Molecular & Cellular Proteomics 2007 : meta-analysis of two affinity purification/mass-spectometry studies
PRS RRS
Collins
CYC2008
PRS RRS SGA
PCC 0.2 PCC 0.3 PCC 0.5
Functional dataset Costanzo et al., Science 2010 : Synthetic Genetic Array (SGA) and
genetic interaction profiles based on PCC values
4 3 2 1 DB-1 AD-2 DB 1 2 AD DB 1 2 AD 2 AD 2 AD DB 1 DB 1 DB 1 2 AD DB 1 DB-1 AD-2 DB 1 2 AD DB-1 DB-1 DB 1 2 AD DB 1 DB 1 DB 1 2 AD DB 1 DB-1 AD-2 DB 1 2 AD 2 AD 2 AD 2 AD DB 1 DB 1 2 AD DB 1 DB-1 AD-2 DB 1 2 AD DB 1 2 AD DB 1 DB-1 DB-1 DB-1 DB-1 AD-2 DB 1 2 AD 2 AD AD-2AD-2 AD-2 AD-2 DB-1 DB-1AD-2 AD-2 AD-2 AD-2 AD-2 AD-2 DB-1 DB-1AD-2 AD-2 AD-2 DB-1 AD-2 AD-2 AD-2 DB-1 AD-2 DB-1 DB-1 DB-1 AD-2 DB-1 AD-2 2AD AD 2μ (pQZ213) (pDestAD)AD Cen DB 2μ (pQZ212) DB Cen (pDestDB) DB 1 2 AD 2 AD 2 AD DB 1 DB 1 DB 1 (2) (4) (3) (1)
Overlap between PRS pairs with the different assay
Assay configuration choosen: DB 2μ - AD 2μ
Pick colonies
Sequence to identify the ORF
Lit−BM−13
Number of PPIs in map (x 1,000 PPIs)
0 2 4 6 8 10 0 0.1 0.2 0.3 0.4 0.5 Fr
action of the estimated full inte
ractome
(Yu et al., Science 2008)
Coverage of binary PPI maps
Union
YI1
Ito-Core
Uetz
Tarrassov
Random Reference Set (RRS)
Positive Reference Set (PRS)
116 interactions with >4 papers
200 pairs extracted from
~18.106 possible pairs from
predicted yeast ORFeome
Recovery rate of co-complex datasets Recovery rate of functional datasets
Binary PPIs within complexes are detected with similar sensitivity.
Pairs with increasingly similar genetic interaction profiles tend to correspond to binary PPIs.
116 PRS Pairs Positive 4 3 2 1 . . . . . . . . . . . . 200 RRS Pairs
Mate every FiPP in quadruplicate + PRS/RRS as controls
Pick colonies
Remove residual autoactivators (AA)
-LWH + 3AT -LH + 3AT + CHX
F
raction of pairs rec
ov
ered (%)
0
10
20
1,286 PPI with confirmed ORF identity + PRS/RRS + Lit-BM-13
Validation by G. princep Protein Complementation Assay (GPCA)
gLuc X Y X Y JAK P P STAT PRS RRS YI2 YI2 1 Screen YI2 2-3 Screens Lit-BM-13
Validation rate of YI2 (11%) as good as Lit-BM-13 (13%) but less than the PRS (23%)
in previously published PPI new from YI2
YI2 contribution to YeRI :
nodes : 21% (467 out of 2,212) edges : 32% (1,198 out of 3,712)
YI2 detects novel interactions compared to the previous binary datasets
All yeast interactome datasets are biased for proteins that are more abundant and more studied. This suggest that part of these biases are biological.
- Validate the YI2 dataset with another orthogonal assay to further asses the quality of our dataset
gPCA
YI3
- Expand the screens to the complete search space for S. cerevisiae, including
candidate proto-genes (~1,139 genes) (Carvunis et al., Nature 2012)
This work was supported by the grant R01HG006061
Motivation
Less than half of the yeast interactome covered
Large contribution from Lit-BM-13 which is sociologically biased. All of the datasets are of good quality when compared to PRS/RRS. PPIs from PCA seem to be of a different nature.
Recovery of binary PPI in protein complexes and functional datasets
Preliminary results
Future plans
YI2 map Union of other Y2H map PCA map (Tarrasov) Non-Binary map Lit-BM-13 Number of literature citations
P ro te in c el lu la r a bundan ce >=100 75 50 25 0 ~4k ~5k DB AD Y2H PCA
Search space and assay
PRS RRS
-LWH + 3AT
-LH + 3AT + CHX
Number of PPI in bin
Alternative validation assay : Mammalian Protein-Protein Interaction Trap
(MAPPIT)
- Perform additional screens of the complete search space to generate a first
Yeast Reference Interactome map (YeRI)
Current map of the yeast interactome
Towards a yeast reference interactome
Despite an increasing number of interactomic datasets already available for model organisms and humans, many aspects remain contradictory, debatable or unclear due to the lack of complete high-quality networks.
For example, does macromolecule connectivity in interactome maps reflect functional importance or sociological biases? Do biophysical interactions always reflect functional relationships?
Here we propose to build a new, alternative view of an interactome map for S.
cerevisiae for which a lot of genome-wide datasets, protein-protein interaction
(PPI) maps and functionnal maps based for instance on expression or genetic interactions, are already available.
Fr
action of pairs rec
ov
ered (%)
Fr
action of pairs rec
ov
ered (%)
Fr
action of pairs rec
ov
ered (%)
Approach to build a reference map
Our stringent pipeline means: -high specificity : no positive
pairs detected in the RRS - low sensitivity, adressed by multiple screens : most positive
where detected in only one screen, but the use of multiple
screens allows maximum
YI2 is a good quality dataset
New nodes
Old nodes New edgesOld edges
0 0.1 0.2 0.3 0.4 0.5 Cen-Cen(1)2μ-2 μ (2) Cen-2 μ (3) 2μ-Cen (4) PRS RRS Fr
action of pairs rec
ov
