1. Protein Targeting by Functional Linkage of Non-Homologous Proteins with examples from M. tuberculosis Genome-wide functional linkage map Structural Genomics of Complexes: Identifying subunits of complexes by analyzing co-evolution of non- homologous proteins, from genome-wide functional linkage maps

2. Limitations of Relying Entirely on Homology-Based Targeting Many (most ?) proteins function in complexes made up of non-homologous proteins Some (many ?) proteins are crystallizable only with their functional partners

3. Limitations of Relying Entirely on Homology-Based Targeting Many (most ?) proteins function in complexes made up of non-homologous proteins Some (many ?) proteins are crystallizable only with their functional partners Suggests that targeting of non-homologus, functionally linked proteins may offer a useful shortcut to learning protein structures and functions

4. Identifying Subunits of Protein Complexes by Analyzing the Co-evolution of Non-homologous Proteins Structural Genomics of Protein Complexes

5. 4 Methods to Infer Non-Homologous Protein Pairs that have Co-evolved and hence are Functionally Linked Rosetta Stone Protein fusion Phylogenetic Profile Protein co-occurrrence Gene neighbor Constant separation Operon Small separation

6. Figure 7. M. Strong, T. Graeber et al.

7. Classical graphical representation of protein functional linkages Research of Michael Strong and Morgan Beeby Requiring 2 or more functional linkages: 1,865 genes make 9,766 linkages Functional Linkages Between Genes of M. tuberculosis

8. Hierarchical Clustering of the Combined Genome-Wide Linkage Map for M. Tb. Reveals Complexes and Pathways Genome-wide functional linkage map based on 4 methods: Clustered linkage map showing complexes and pathways: Cluster similar linkage patterns  ach cluster is a complex or pathway

9. Detoxification Polyketide and non-ribosomal peptide synthesis Energy Metabolism, oxidoreductase Deg. of Fatty Acids Virulence Energy Metabolism, oxidoreductase Amino acid Biosynthesis Emergy Metab. Respiration Aerobic Lipid Biosynthesis Degradation of Fatty Acids Amino Acid Biosynthesis (Branched) Synthesis and Modif. Of Macromolecules, rpl,rpm, rps Biosynthesis of Cofactors, Prosthetic groups Purine, Pyrimidine nucleotide biosynthesis Novel Group Sugar Metabolism Aromatic Amino Acid Biosynthesis Energy Metabolism, Anaerobic Respiration Two component systemsCell Envelope Cytochrome P450Chaperones Biosynthesis of cofactors Cell Envelope, Cell Division Transport/Binding Proteins Energy Metabolism TCA Broad Regulatory, Serine Threonine Protein Kinase Cell Envelope, Murein Sacculus and Peptidoglycan Transport/Binding Proteins Cations Energy Metabolism, ATP Proton Motive force Fig 4. M. Strong, T. Graeber et al.

10. Quantitative Assessment of Inferred Protein Complexes

11. Calculating Probabilities of Co-evolution Phylogenetic Profile Rosetta Stone Gene Neighbor Operon N= number of fully sequenced genomes n= number of homologs of protein A m = number of homologs of protein B k = number of genomes shared in common X= fractional separation of genes n = intergenic separation

12. Combining Inferences of Co- Evolution from 4 Methods We use a Bayesian approach to combine the probabilities from the four methods to arrive at a single probability that two proteins co-evolve: where positive pairs are proteins with common pathway annotation and negative pairs are proteins with different annotation

13. Benchmarking this Approach Against Known Complexes Ecocyc: Karp et al. NAR, 30, 56 (2002) True positive interactions are between subunits of known complexes and false positive ones are between subunits of different complexes. For high confidence links, we find 1/3 of true interactions with only one 1/1000 of the false positive ones Random

14. Benchmarking our Approach Against Known Complexes True positive interactions are between subunits of known complexes and false positive ones are between subunits of different complexes. For the first few hundred pairs of high confidence links, about 50% are between subunits of known complexes

15. Example Complex: NADH Dehydrogenase I 11 of 13 subunits detected

16. Example Complex: NADH Dehydrogenase I 11 of 13 subunits detected 3 false positives

17. Assessing Inferred Linkages for M. Tb Genome (Michael Strong, 2003) Accuracy Conclusions: 100 bp operon threshold is adequate A functional linkage by 2 or more methods is reasonably accurate

18. CtaD CtaECtaC Functional Linkages Among Cytochrome Oxidase Genes CtaB Functional linkages relate all 3 components of cytochrome oxidase complex and also CtaB, the cytochrome oxidase assembly factor These genes are at four different chromosomal locations Membrane proteins linked to soluble proteins

19. From Inferred Protein Complexes to their Structures

20. PE, PE-PGRS, and PPE Proteins in M. tuberculosis 38 PE proteins; 61 PE-PGRS proteins; 68 PPE proteins Together compromise about 5 % of the genome No function is known, but some appear to be membrane bound No structure is known: always insoluble when expressed Goal: use functional linkages to predict a complex between a PE and a PPE protein: express complex, and determine its structure Research of Shuishu Wang and Michael Strong The Problem of PE and PPE Proteins in M. tb

21. Construction of a co-expression vector to test for protein-protein interactions (Mike Strong) pET 29b(+) T7 promoter lac oper. RBS Nde1 HindIIIKpn1NcoI RBS gene A gene B Thrombin site His tag polycistronic mRNA transcription translation protein A protein B (with His tag) If proteins interact (protein-protein interaction) If proteins do not interact

22. When co-expressed, the PE and PPE proteins, inferred to interact, do form a soluble complex, Mr = 35,200 Sedimentation equilibrium experiments: Rv2430c + Rv2431c fraction 49, in 20mM HEPES, 150mM NaCl, pH 7.8 Concentration OD 280 0.7, 0.45, 0.15 Expected Mr: Rv 2431c (PE) 10,687 (10563.12 from Mass Spec) Rv2430c+His tag (PPE) 24,072 (23895.00 from Mass Spec) Possibly suggests a 1:1 complex between these two proteins

23. Crystallization trials of the Complex Between PE Protein Rv2430c and PPE Protein Rv2431c

24. Summary Many functional lnkages are revealed from genomic data (high coverage)

25. Summary Many functional lnkages are revealed from genomic data (high coverage) Clustered genome-wide functional maps can reveal and organize information on complexes (and pathways)

26. Summary Many functional lnkages are revealed from genomic data (high coverage) Known subunits of E. coli complexes can be identified with high accuracy from functional linkages Clustered genome-wide functional maps can reveal and organize information on complexes (and pathways)

27. Summary Many functional lnkages are revealed from genomic data (high coverage) Known subunits of E. coli complexes can be identified with high accuracy from functional linkages Clustered genome-wide functional maps can reveal and organize information on complexes (and pathways) A protein complex suitable for structural studies has been revealed from functional linkages

28. Summary Many functional lnkages are revealed from genomic data (high coverage) Known subunits of E. coli complexes can be identified with high accuracy from functional linkages Clustered genome-wide functional maps can reveal and organize information on complexes (and pathways) A protein complex suitable for structural studies has been revealed from functional linkages The procedures for identifying and producing protein complexes can be adapted for high thruput

29. Protein Interactions in M. tb. Analysis of M.tb. Genome Michael Strong, Debnath Pal, Sulmin Kim Whole Genome Interaction Maps Michael Strong, Tom Graeber, Huiying Li, Matteo Pellegrini Methods of Inferring Interactions Edward Marcotte, Matteo Pellegrini, Todd Yeates, Michael Thompson PI of Tb Structural Genomics Consortium Tom Terwilliger