Proteinprotein interactions Courtesy of Sarah Teichmann Jose B
Protein-protein interactions Courtesy of Sarah Teichmann & Jose B. Pereira-Leal MRC Laboratory of Molecular Biology, Cambridge, UK EMBL-EBI
Stable vs. transient protein-protein interactions Stable complex Transient Interaction Stable complex: homodimeric citrate synthase Transient Signaling Complex Rap 1 A – c. Raf 1 Interface 4890 Å2 Hydrophobic interfaces Interface 1310 Å2 “Hydrophilic” interfaces Multi-domain protein
Stable vs. transient protein-protein interactions
Interface Constraints in Multi-domain Proteins No. proteins in S. cerevisiae: Average identity S. cerevisiae. S. pombe: 1844 42% 572 43% 186 45%
Summary • Sequence conservation: Stable complexes > transient > other • Co-expression/co-regulation: Stable complexes > transient
Using publicly available interaction data Are there know interaction partners for you pet protein? Check if: 1. There are interactors for your protein in the literature 2. There are databases of interactions where your protein may appear 3. There are homologues of your protein in the protein interaction databases 4. You can predict interactors by other means? 5. This failing, at this point you go back to the bench…
Using publicly available interaction data 1. Are there interactors for my protein in the literature ? Problems: • Low coverage • Does not include results from high throughput experiments • Gene names may not be consistent
Using publicly available interaction data 2. Are there databases of interactions where my protein may appear? Some DBs: BIND, MINT (General) + organism specific databases (e. g. MIPS/CYGD) Caution! Check: -the experimental methods used to identify the interaction (e. g. high error rate in large scale yeast-two hybrids) -check the method used to incorporate the interaction in the database (e. g. manual curation vs. literature mining using “intelligent” algorithms)
Using publicly available interaction data 3. Are there homologues of my protein in the protein interaction databases? We are assuming that protein interactions are conserved in evolution • Plenty of evidence that they are… • BUT, how do you define homologous/orthologous ? Make sure that you understand the limits of such “prediction”: two single-gene family products interact in one organisms, and they also exist as single gene-family products in another genome --> potentially good prediction -but the original interactions was identified in a large scale Y 2 H, is not supported by any other observation and one of the proteins has 133 described interactors in that experiment… --> likely a false positive (you learned nothing about your protein!)
Apparently no one in the world ever bothered to look at your favorite protein… now what? Computational Prediction of protein interactions (functional associations) Caution: • Computational methods are good at finding functional associations • A functional association is not the same thing as a physical interaction • Since : • we don’t know how many of the experimentally derived interactions are true/biologically significant • We don’t know how many interactions exist • Impossible to determine how good predictions REALLY are (this becomes more important as the number of predictions you make increase [automation])
Experimental techniques Yeast two-hybrid screens MS analysis of tagged complexes Correlated m. RNA expression levels Synthetic lethality
m. RNA expression level (ratio) Experimental techniques Yeast two-hybrid screens MS analysis of tagged complexes Correlated m. RNA expression levels Synthetic lethality Time-> Predict Close Functional relationship. Interaction from 18 M of Microarray timecourse of Unknown Member of from Cluster (2 Random Interacting Ribosomal Proteins) relationship 18 M 1 ribosomal protein
Experimental techniques Yeast two-hybrid screens MS analysis of tagged complexes Correlated m. RNA expression levels Synthetic lethality
How good is the data? (von Mering et al. , Nature 417: 399)
How good is the data? (von Mering et al. , Nature 417: 399)
How good is the data? (von Mering et al. , Nature 417: 399) ”We estimate that more than half of all current high-throughput interaction data are spurious”
Computational prediction of protein interactions Gene fusion events Tryptophan synthetase a b fusion Trp. C Trp. F 1 PII Enright et al (1999) Nature 409: 86 Marcotte et al (1999) Science 285: 751 Fused in E. coli Unfused in some other genomes (Synechocystis sp. and Thermotoga maritima. )
Computational prediction of protein interactions Phylogenetic profiles Pellegrini et al (1999) PNAS 96: 4285
Computational prediction of protein interactions Conservation of Gene neighborhood e. g. operons in bacteria Not really applicable to eukaryotes, except, to some extent, C. elegans However, there is hope for eukaryotes: -adjacent genes are frequently co-expressed (co-regulated) -co-regulated proteins are likely to be functionally associated maybe this principle may be used for prediction of interactions
Computational prediction of protein interactions Mirror trees Proteins that physically interact tend to co-evolve Pazos and Valencia (2001) Protein eng. 14: 609
Computational prediction of protein interactions Pre-computed predictions: where to find them?
The Big Picture
Scale-free networks (Ravasz et al. , Science 297: 1551)
Some examples of systems with a scale invariant organization Broccoli World-wide web (some) Food webs Roads Social networks
Scale free behavior in protein interaction networks: scale free or scale invariance self-similarity Scale invariance gives insight into robustness of biological systems
Modular networks
Hierarchical networks
Identification of functional modules from protein interaction data Messy data Functional modules Graph theory formalisms Pereiral-Leal, Enright and Ouzounis (2003) Proteins in press Custering
From functional modules to pathways Canonical pathways Pereiral-Leal, Enright and Ouzounis (2003) Proteins in press
So. . You know your two proteins interact… do you want to know how? Prediction of the molecular basis of protein interactions
Molecular basis of protein interaction “Tree determinant residues” Rab REP MSA REP x Ras Rho Arf Ran Prediction _ Experimental tests Pereira-Leal and Seabra (2001) J. Mol. Biol. Pereira-Leal et al (2003) Biochem. Biophys. Res. Com. +
Molecular basis of protein interaction “Tree determinant residues” Continued… Sequence Space algorithm AMAS (part of a bigger package) Casari et al (1995) Nat. Struct. Biol 2(2)
Molecular basis of protein interaction In silico docking Requires 3 D structures of components Conformational changes cannot be considered (rigid body)
- Slides: 36