B-DEBATE - THE FUTURE OF PLANT GENOMES. HARVESTING GENES FOR AGRICULTURE; 9-11 October 2012, CRAG Barcelona

1. THE COMPLEXITY OF PLANT GENOMES
Genome structure, gene functions and beyond
Klaas Vandepoele
Barcelona, October 10th 2012
Department of Plant Biotechnology and Bioinformatics, Ghent University
Department of Plant Systems Biology, VIB - Belgium
http://twitter.com/plaza_genomics

2. OVERVIEW
 And then there were many: plant genome sequences
 PLAZA: a web-based plant comparative genomics
toolbox
 Genome organization and evolution
 The quest for plant orthologous genes
 Unravelling gene functions using integrative plant
genomics
 Cross-species gene function analysis

3. 1. OVERVIEW PLANT GENOME SEQUENCING
Individual
institutes
International
consortia
Today: ~40 (complete) plant genome sequences

5. GENOME ANNOTATION
Functional
Annotated
Genoscope BGI JGI EST genes
Genomic DNA
Sequences
Downstream
analysis
Artemis Manual
GenomeView Curation
Coding potential Repeats
search Training
set
Intron potential Build splice
IMM SpliceMachine
search Site models
Repeat
Mask
Intergenic
potential search
Automatic Mask
Eugene repeats
annotation
GenomeView Bogas
tBlastx Blastx Blastn
Expert
Structural
annotation
annotated
genes
Related Swissprot EST
genomes Nr_prot cDNA Gene
Ontology
Functional
annotation
InterPro
Predicted
genes
Source: P. Rouzé

6. EXPLOITING GENOME INFORMATION
 Centralized infrastructure
 Detailed gene catalog per species
 Structural annotation (gene models, UTRs)
 Functional annotation (experimental, sequence-based, systems
biology)
 Intuitive & advanced data mining tools for non-expert
users
 Gene function
 Genome organization
 Pathway evolution
 Data manipulation
 Computational resources

7. Gene family analysis
Genome analysis
>20 tools available
Proost et al., Plant Cell 2009; Van Bel et al., 2012

8. HOMOLOGOUS GENE FAMILIES
>780K proteins
from 25 species
Protein clustering
Phylogenetics
18K trees incl. 420K 22K multi-species gene families
annotated tree nodes covering 83% of the total proteome

9. GENE COLINEARITY & GENOME ORGANIZATION
Chromosome 1
• Represent chromosomes as
sorted gene lists
Chromosome 2
• Identify all homologous gene
pairs between chromosomes (all-
against-all BLASTP).
• Score pairs of homologues in
matrix
1
Gene Homology Matrix (GHM)
i-ADHoRe 3.0 2

10. GENOMIC PROFILES
pairwise
multiple
Simillion et al. (2004) Genome Res. 14, 1095-1106

11. IMPROVED SENSITIVITY TO DETECT DEGENERATE
GENOMIC HOMOLOGY
(#homologous segments)
Proost, Fostier … & Vandepoele, NAR 2011

12. I-ADHORE 3.0
Speed & memory footprint
Fostier, … & Vandepoele, Bioinformatics 2011
Proost, Fostier … & Vandepoele, NAR 2011

13. GENOME-WIDE COLINEARITY
Z. mays WGDotplot
O. sativa

14. MULTI-SPECIES COLINEARITY
profile

15. WHOLE-GENOME CIRCULAR DOTPLOT
Reference: O. sativa
Inner circle: duplicated regions
Outer circle: inter-species colinear regions

16. Gene family analysis
Genome analysis
Proost et al., Plant Cell 2009; Van Bel et al., 2012

17. FUNCTIONAL ANALYSIS OF SPECIES-SPECIFIC
GENE DUPLICATES
Species
specific
duplicates
Divide in block & tandem
duplicates
Gene-sets
PLAZA
workbench
GO enrichment
Proost et al., Plant Cell 2009

18. FUNCTIONAL ANALYSIS OF SPECIES-SPECIFIC
GENE DUPLICATES
Species
specific
duplicates
Divide in block & tandem
duplicates
Gene-sets
Gene Ontology
PLAZA
workbench
GO enrichment

19. FUNCTIONAL ANALYSIS OF SPECIES-SPECIFIC
GENE DUPLICATES
Species
specific
duplicates
Divide in block & tandem
duplicates
Gene-sets
Gene Ontology
PLAZA
workbench
GO enrichment

20. CORE HISTONE CLUSTERS IN C. REINHARDTII
Synteny plot
Proost et al., Plant Cell 2009

21. THE QUEST FOR PLANT ORTHOLOGS
 Plants are paleopolyploids
 Dynamic genome organization
 Large fraction of multi-gene
families
 Absence of simple 1:1
orthology relationships

22. Source: Y. Van de Peer

23. GENE DYNAMICS IN THE GREEN LINEAGE
Green algae Brown algae Land plants
Diatoms

24. PLANT GENE FAMILIES, A TALE OF DUPLICATIONS
F-box protein domain gene family

25. PLAZA INTEGRATIVE ORTHOLOGY VIEWER
•Tree-based orthologs (TROG) inferred using tree reconciliation
•Orthologous gene families (ORTHO) inferred using OrthoMCL
•Anchor points refer to gene-based colinearity between species Van Bel et al.,
•Best hit families (BHIF) inferred from Blast hits including inparalogs Plant Physiology 2012

26. COMPLEX GENE ORTHOLOGY RELATIONSHIPS
Query species: A. thaliana
Target species

27. 3. PLANT –OMICS SPACE
Mochida and Shinozaki, 2011

28. INTEGRATIVE PLANT GENOMICS
 Explore genome-wide –omics data sets to study gene function
and regulation
 Transcriptomics (Microarrays|RNA-Seq)
 Interactome data (Y2H|TAP)
 Regulatory interactions (TF|miRNA-target|TF motifs)
 Include expert gene annotations
 Dedicated databases (e.g. phenotypes, metabolomics)
 Text-mining

29. GENE NETWORK ANALYSIS
 Features
 Integration heterogeneous –omics data sources
 Different gene-gene associations with varying quality
 Missing data
 Exploit network-guided guilt-by-association
principle
 Methodologies
 Simple un-weighted/weighted graphs
 Probabilistic models
Lee et al., 2010

30. EXPERIMENTAL ARABIDOPSIS GENE-GENE
ASSOCIATION DATA
Datatype # Genes # Associations (% unique) Source
PPI 3,194 7,210 (75%) CORNET
AraNet* 19,647 1,062,222 (99%) Lee et al., 2010
TF targets 9,422 13,037 (99%) AtRegNet (AGRIS)
GO 6,588 89,100 (n.a.) GeneOntology.org / TAIR
Total 22,492 1,089,661
* Probabilistic network integrating heterogeneous genomic features
Research objectives:
• Infer functional gene modules starting from experimental data
• Identify regulatory properties of genes, modules and network
• Explore cross-species functional annotation
Heyndrickx and Vandepoele, 2012

31. DELINEATING ARABIDOPSIS GENE MODULES
 Transform gene-gene associations in networks and
functional gene modules

32. CONVERTING STATIC GENE ASSOCIATIONS INTO
FUNCTIONAL EXPRESSION MODULES
 Classical approach
1. Clustering expression data
• Guide-gene (gene-centric)
• Non-targeted (global)
2. Functional analysis modules using
enrichment statistic
 Challenges - weaknesses
 Which microarray samples to include?
 Functional information integrated a
posteriori
Aoki et al., 2007

33. EXPRESSION-BASED CLUSTERING
 Integrate a priori functional information
during module detection
 Semi-supervised clustering strategy
considering multiple query genes and
multiple expression compendia
 Rank aggregation through scoring
function
 maximize coexpression towards multiple seeds
showing dynamic expression profile

34. ANALYSIS GENE MODULES

35. PROPERTIES INPUT – MODULE DATA
 40% of the genes in the modules is present in more than one input data type
 only 3% of the gene pairs within a module having support by more than one
primary data type

36. MODULE OVERLAP
Primary Data Modules
Datatype # Genes # Associations (% # Genes # Modules Functional Motif
unique) (% unique) Enrichment Enrichment
PPI 3,194 7,210 (75%) 597 72 (95%) 51 43
AraNet 19,647 1,062,222 (99%) 6,377 419 (99%) 116 172
TF targets 9,422 13,037 (99%) 5,127 518 (96%) 51 224
GO 6,588 89,100 (n.a.) 7,750 1,105 (99%) 943 341
Total 22,492 1,089,661 13,428 2,114 1,161
Non-redundant 13,142 1,563 676 772
Modules
 >99% modules found through a single input data type

37. FUNCTIONAL AND CIS-REGULATORY COHERENCE
OF PLANT MODULES
Cis-regulatory element analysis
• Weeder / MotifSampler de novo
motif finding (1544 motifs)
• Overlap with known plant motifs
AGRIS/PLACE (34%)
Functional enrichment analysis
• Over-representation hypergeometric
distribution + FDR
• Non-electronic GO annotations +
embryo-lethal gene (SeedGenes)
 40% of the modules could be linked to a significant functional enrichment (GO BP -
embryo lethality)
 98% of the modules have 1 (or more) gene(s) with a known experimental annotation

38. FUNCTIONAL MODULE REPERTOIRE
http://bioinformatics.psb.ugent.be/cig_data/plant_modules/

39. CROSS-SPECIES MODULE ANALYSIS
Affymetrix GeneChip
NCBI Gene Expression Omnibus
1563 Arabidopsis Integrative
modules orthology

40. CONSERVED MODULE EXPRESSION COHERENCE
Lipid biosynthesis
 58% of modules shows significant coexpression coherence (3 or more species)
 >43,000 unknown genes from 6 other plants receive module-based functional
annotations

41. MODULE-BASED FUNCTION PREDICTIONS
 Can we recover new experimental Arabidopsis gene – GO BP
annotations?
Data freeze Evaluation
1460 Arabidopsis genes with predictions receive new exp. GO-BP
Unknown Unknown Exp. BP Other Exp. BP Total
#Pred. #Conf #Pred #Conf #Pred #Conf #Pred # Conf
All Genes 197 75 (38.1%) 255 108 (42.4%) 1,008 251 (24.9%) 1,460 434 (29.7%)
Conserved 166 65 (39.2%) 195 80 (41%) 871 215 (24.7%) 1,232 360 (29.2%)
Not Conserved 48 10 (20.8%) 83 31 (37.3%) 315 52 (16.5%) 446 93 (20.9%)

42. DNA ENDOREDUPLICATION
• PPI module: predicted to be
involved in DNA endoreduplication
• Experimental validation shows that
AT1G06590 T-DNA shows
perturbed endoreduplication index
(Quimbaya et al., 2012)
Plant Mutants Flow Cytometry
Quimbaya, Vandepoele,… De Veylder, 2012

43. 4. INTEGRATED CO-EXPRESSION – ORTHOLOGY
NETWORKS
Movahedi et al., 2012

44. 3-WAY SPECIES CO-EXPRESSION COMPARISON FOR ETG1
• Conserved DNA
replication module
• Conserved E2F target
gene (TTTCCGC)
• Role in sister
chromatin cohesion
Movahedi et al., 2012; Takahashi et al., 2010

45. SORTING OUT PLANT (CO-)ORTHOLOGS USING
EXPRESSION CONTEXT CONSERVATION
Protein integrative orthology
Expression Context Conservation
scores (p-value < 0.05)
Inparalogs (species-specific
duplicates)

46. 4. CONCLUSIONS
 Need for advanced & user-friendly tools to characterize new genomes
 Complexity and quality genome sequences
 Scalability with increasing number of genomes
 Integrative approaches combining multiple methods outperform individual
methods* and provide users a more complete view
 Computer power
 Visualization
 Large discrepancy in the functional gene associations between the different
experimental data sets
 A large fraction of the module-based functional predictions are biologically
valid and can be transferred across species
 Comparative network approaches provide a powerful tool to integrate
functional genomics data
* Quest for Orthologs Consortium, Bioinformatics 2012

47. ACKNOWLEDGEMENTS
 Ken Heyndrickx
 Michiel Van Bel
 Sebastian Proost
 Sara Movahedi
 Mauricio Quimbaya

48. ACKNOWLEDGEMENTS
Further reading
 Proost, S., Van Bel, M., Sterck, L., Billiau, K., Van Parys, T., Van de Peer, Y., and
Vandepoele, K. (2009). PLAZA: a comparative genomics resource to study gene and
genome evolution in plants. Plant Cell
 Heyndrickx, K.S., and Vandepoele, K. (2012). Systematic identification of functional
plant modules through the integration of complementary data sources. Plant Physiol.
 Movahedi, S., Van Bel, M., Heyndrickx, KS., Vandepoele, K. (2012) Comparative co-
expression analysis in plant biology. Plant, Cell & Environment