1. 6/26/2015 Function Prediction of Protein Complexes with Domain Correlation Ya Zhang Xue-Wen Chen University of Kansas Introduction  Protein complexes: groups of proteins that interact/associate with each other.  Many molecular machineries are protein complexes  e.g., proteasome, responsible for degradation of unwanted proteins, consists of more than 50 proteins Motivation  Previous efforts in analyzing protein complex data were largely based on the protein constitution of the complexes.  Studying complexes at protein level does not always reveal the function linkage among complexes.  The cytoplasmic ribosomal large subunit vs. the mitochondrial ribosomal large subunit  In this study, we analyze the domain composition of protein complexes to predict the functions of protein complexes. Discover domain functional modules  Support of a set of domains X={d 1, d 2,.., d k }, is the fraction of protein complexes containing X. supp(X)=N X /N  H-confidence of X is: hconf(X)=supp(X)/max i (supp(p i ))  A set of domains X ={d 1, d 2,.., d k } is a hyperclique pattern if hconf(X) ≥ h c and supp(X) ≥ s.  Understanding the composition and function of protein complexes is an important research focus in biological science. Domaincytoplasmicmitochondrial Ribosomal_L5YGR085c; YPR102cYDR237w Ribosomal_L23YOL127wYDR405w Ribosomal_L11YDR418w; YEL054cYNL185c Ribosomal_L2YFR031c-a; YIL018wYEL050c Ribosomal_L6YGL147c; YNL067wYHR147c  The cytoplasmic ribosomal large subunit vs. the mitochondrial ribosomal large subunit:  Both function for protein synthesis  Share no single protein but 14 domains Method  Domain functional modules  Set of domains that perform some elementary functions in protein complexes.  Extensively shared among protein complexes. Domain-domain network Nodes: domains Edges: shared membership High-throughput Protein Complexes  High-throughput experiments have produced a large amount of protein complex data.  Gavin, et al. (Nature, 2002)  TAP : Tandem Affinity Purification  Ho, et al. (Nature, 2002)  HMS-PCI: High-throughput Mass Spectromic Protein Complex Identification  But the biological functions of the protein complexes are largely unknown. ComplexIDFunction(Pred.)MIPS annotation 550.1.108 Protein synthesisprobably protein synthesis turnover 550.1.104 Protein synthesisprobably protein synthesis turnover 550.1.140; 550.1.142; 550.1.155 DNA replicationprobably RNA metabolism 550.1.39; 550.1.42 oxidoreductaseprobably intermediate and energy metabolism 550.2.56 DNA replicationN/A 550.1.190 550.1.211; 550.1.220; 550.1.205; 550.1.228 DNA replicationprobably transcription/DNA maintanance/chromatin structure Domain Functional ModulesProtein Complexes RNA_pol_Rpb1_1; RNA_pol_Rpb1_2; RNA_pol_Rpb1_3; RNA_pol_Rpb1_4; RNA_pol_Rpb1_5; RNA_pol_L; RNA_pol_A_bac; RNA_pol_Rpb2_1; RNA_pol_Rpb2_2; RNA_pol_Rpb2_3; RNA_pol_Rpb2_5; RNA_pol_Rpb2_6; RNA_pol_Rpb2_7; RNA_pol_Rpb5_N; RNA_pol_Rpb5_C; DNA_RNApol_7kD; RNA_pol_N; RNA_pol_Rpb8; RNA_pol_Rpb6 RNA polymerase I; RNA polymerase II; RNA polymerase III Ribosomal_L14; Ribosomal_L1; Ribosomal_L5; Ribosomal_L5_C; Ribosomal_L23; Ribosomal_L6; KOW; Ribosomal_L11_N; Ribosomal_L11; Ribosomal_L2; Ribosomal_L2_C; L15; Ribosomal_L3; Ribosomal_L13 cytoplasmic ribosomal large subunit; mitochondrial ribosomal large subunit Ribosomal_S17; Ribosomal_S9; S4; Ribosomal_S5; Ribosomal_S5_C; Ribosomal_S14; Ribosomal_S10; Ribosomal_S15; Ribosomal_S2; Ribosomal_S7; Ribosomal_S19 mitochondrial ribosomal small subunit; cytoplasmic ribosomal small subunit Transket_pyr; Pyr_redox_2; Pyr_redox; Biotin_lipoyl; 2- oxoacid_dh; E1_dh; GIDA; Pyr_redox_dim Pyruvate dehydrogenase; 2-oxoglutarate dehydrogenas Results: hypercliques Results: function prediction Discover domain functional modules  Discover domain functional modules as cliques in domain-domain network Protein Complexes (Higher order functions) Domain Functional Modules (Elementary functions) C1 C2 C3 M1 M2  Two criteria  Frequency of occurrence  Correlation / association  The modules indicate the functions of the corresponding protein complexes.  Identify domain functional modules as hyperclique patterns (Xiong et al. 2005).