Genome-wide multilevel spatial interactome model of rice

Transcription

1 Sino-German Workshop on Multiscale Spatial Computational Systems Biology, Beijing, Oct 8-12, 2015 Genome-wide multilevel spatial interactome model of rice Ming CHEN ( 陈铭 ) mchen@zju.edu.cn College of Life Sciences Zhejiang University

2 Integrative Bioinformatics Omics Genomics Transcriptomics Proteomics Metabolomics Phenomics Integrative Bioinformatics Sequence annotation Gene ontology Gene coexpression Protein interactions Pathway information Trait corelation? 1. Confirmation Plant Transcriptomics (non-coding RNAs) 2. Genome-wide multiple scale biological network Mutants Haplotypes Overexpression Complementation Transgenics Knockout BSA Function

3 Background Genome-wide Gene regulatory network of rice PNAS. 2011, Insuk Lee et. al., , vol. 8.

4 Our GRNs Gene regulatory network construction 1) Identify the functional associations: a: Gene co-expression b: Gene neighborhoods c: Phylogenetic profile d: Genetic interaction e: Domain co-occurrence 2) Distinguish PPI from gene regulatory network using a probabilistic model proposed by Naoki Nariai et al

5 Background Proteomics Structural, Expressional, Functional Proteomics proteins in Uniprot Exp.Method Proteins X-RAY NMR 9619 ELECTRON MICROSCOPY 569 HYBRID 70 other 165 Total

6 Rice Proteomics

7 OsPAD: Proteome Annotation Database of 2D-PAGE Biochip J, 2009

8 Annotations RPD annotated spot KEGG annotation OsPAD annotation

9 Virtual Gel Normal gel Gel comparison

10 Protein expressed patterns Current Protein and Peptide Science, 2011

11 PRIN: a predicted rice interactome network BMC Bioinformatics. 2011

12 Framework

13 Interolog gene neighbourhood gene fusion domain fusion gene co-expression phylogenetic profile subcellular co-location domain interaction GO similarity

14 PRIN: example

15 PRIN: example

16 Most evolutionary conserved interactions Genomics, Proteomics & Bioinformatics, 2011

17 Application

18 Background Genome-wide metabolic network of plants The metabolic reconstruction of rice is still to be explored!

19 MyBioNet Core metabolic network of rice includes 3035 nodes, 5774 edges Bioinformatics. 2011

20 Problem: subcellular location

21 Background Subcellular Localization To understand protein function: location link to its function Is a critical step in genome annotation Experimental methods can be time & money costing Computational methods turn out to be effective however show bias More applications: improve drug target identification (e.g. secreted, plasma membrane protein) In animals, aberrant subcellular localization of proteins lead to diseases (e.g. cancer, Alyheimer s disease)

22 Our Approach Collect individual predictor for subcellular location Collect gold standard Evaluation of predictors Predictors selection Develop integrated methods

23 Predictors collected -Web Servers or standalone software that can predict subcellular locations, signal peptide and location-specific proteins were benchmarked (7 standalone) Predictors were collected from published articles, only predictors still work were kept for further study. 2. Free web servers support batch search and allow results grab were picked, others were removed. 3. Standalone software and programs were installed and ran locally. 4. At last 19 predictors were used for integrated methods developing. Predictors Status Predictor Locations Use for integration WolfPSORT pero,cyto,nuc,mito Y PredSL mito Y NucPred nuc Y.. WegoLoc golg,nuc,cyto,plas N

24 Gold standard collected - Datasets were built for both training and test. SUBA3 and PPDB Step1. We downloaded and processed non-isoform fungal sequences that have experimentally verified subcellular locations from the newest Uniprot version (13791 sequences collected). Step2. We used CD-HIT to remove redundant proteins (12399 sequences remain). Step3. We chose 10 locations for our work, and mapped the Uniprot proteins to these 10 locations (12399 sequences remain). Step4. We removed proteins that not present in these locations (10904 sequences finally got).

25 Methods Subcellular location prediction PLoS One, 2013

26

27 PSI: A comprehensive and integrative approach for accurate plant subcellular localization prediction

28 Organelle-focused proteomes and interactomes in rice A total of 55,342 proteins (accounting for 83.42% of the whole rice proteome) obtained their subcellular localizations. the organellefocused interactome involving 36,812 pairs of proteinprotein interactions and 5,092 proteins was located in nine organelles.

29 Function enrichment analysis a total of 45 function groups were identified.

30 The sub-network involving coatomer protein subunits triangle motif might be a preferred organization form for proteins to implement biological functions. propose that EMB30 might cowork with coatomer in retrograde Golgi-to-ER transport of dilysine-tagged protein. By gtriescanner

31 Number of Reactions Number of Genes/Reactions 1145 / / / / / / 114 RiceCyc Uniprot KEGG (1928) (412) Brenda (1305) (259) 21 (1375) (244) 260 (6) 7 (407) (10) (58) (5) 54 (755) (63) (45) 26 (21) (32) (8) (70) (11) 0.5 Metabolites:2876 (Reactions:3055) 0.0 EC1 EC2 EC3 EC4 EC5 EC6 Connectivity of metabolites in Rice 1000 Metabolite Absolute Relative(%) Proton % NAD(P) % NADPH % ATP % CO % Diphosphate % 200 NADH % ADP % Number of Metabolites Phosphate % NH % Glutamate % Unpublish

32 Number of Metabolites (or Reactions) Percentage of regulations Percentage of Proteins Percentage of Proteins 0% 17% >50% 33% a Compartmennt 2 Compartment Unique to Compartment b Cytosol (14830 / 1591) Chloroplast (19929 / 2296) Endoplasmic Reticulum (595 / 113) c Vacuole (1303 / 250) Mitochondria (15280 / 693) Nucleus (17648 / 1917) Extracellular (5143 / 484) Peroxisome (361 / 57) Plasma Membrane (7508 / 1996) Golgi Apparatus (154 / 61) 3+ Compartmennt 2 Compartment Unique to Compartment Metabolites Reactions d Nucleus Cytosol Mitochondria Chloroplast Peroxisome Golgi Apparatus Plasma Membrane Vacuole Extracellular Endoplasmic reticulum Nucleus Cytosol Chloroplast Mitochondria Peroxisome Golgi Apparatus Plasma Membrane Vacuole Extracellular Endoplasmic reticulum Gene-gene regulations Protein-protein interactions Gene-metabolite associations Cytosol Chloroplast Mitochondria Extracellular Plasma Membrane Nucleus Peroxisome Endoplasmic Reticulum Vacuole Golgi Apparatus 0.00 Cytosol Nucleus Chloroplast Mitochondria Extracellular Endoplasmic Reticulum Plasma Membrane Vacuole Peroxisome Golgi Apparatus

33 Multi-level Network of Rice Integrate PPIs and Gene Regulations into Metabolic network Genome-wide multi-level network reconstruction Extract protein-protein interactions from public biology database (e.g.: BIND, PlaPid, PRIN, etc.). Integrate the gene regulations into the metabolic network. Collect microrna-target interactions into the metabolic model. Database construction and Network visualization.

34 mrna/mrna* network BiB 2011 Bioinformatics 2014 RNA Biology 2011 ncrna biogensis Briefings in Bioinformatics 2013 NAR 2012 Specific & crosstalk Bioinformatics 2010 RNA Research 2015 RNA 2011 Construction of gene regulatory networks mediated by vegetative and reproductive stage-specific small RNAs in rice (Oryza sativa) PLoS One 2012 J Genetics & Genomics 2013 New Phytologist 2013

35 Multi-level Network of Rice Integrate PPIs and Gene Regulations into Metabolic network Genome-wide multi-level network reconstruction Extract protein-protein interactions from public biology database (e.g.: BIND, PlaPid, PRIN, etc.). Integrate the gene regulations into the metabolic network. Collect microrna-target interactions into the metabolic model. Database construction and Network visualization.

36 Home page Organelle-focused Multiple Level Network

37 3D visualization

38 CELLmicrocosmos

39 Conclusion Proteomics annotation systems for rice. PRIN, PPI networks of rice were constructed. PSI, a highly accuracy web server for plant subcellular localization prediction, was developed. The genome wide multiple level network of rice was reconstructed. sirna/mirna regulatory networks were investigated. A 3D network visualization tool was developed. RiceNetDB are developing for biomolecular regulatory analysis and gene-metabolite mapping. A long way to goooooo collaborations needed!

40 PSI 2.0 Future work Construct the genome scale multiple level network for tissuespecific and context specific Develop algorithms for multiple level network analysis a: network properties analysis b: module identification c: sub-network analysis d: network comparison Phenotype associated network module identification and dynamic analysis based on genome scale multiple level network Deeply annotate the multiple level networks based on a variety of bioinformatics methods such as homology modeling, protein three-dimensional structure docking and so on. Virtual cell model and its applications.

41 Acknowledgements Lili Liu, Yijun Meng, Dijun Chen, Chaogang Shao, Pengcheng Zhu, Haibin Gu Prof. Ralf Hofestädt Dr. Björn Sommer Lu Zhu