ENCORE Evaluated NonCoding Region database

1
ENCOREEvaluated Non-Coding Region database
CODE
TransCODE

• Elia Stupka - elia_at_tigem.it
• Telethon Institute of Genetics and Medicine

2
My background

• Part of the Ensembl team (EBI) annotating the
  human genome, Nature 2001
• Managed the team in charge of the Fugu genome
  annotation in Singapore
• Designed and developed BioPipe, software for
  large scale protocol-driven analysis
• Now working on comparative genomics

3
Genome annotation where are we standing?

• Large effforts have gone into annotating genes
  and gene-related features
• Enough known genes to spawn reliable gene
  prediction algorithms
• Large collections of ESTs and cDNAs to refine
  gene structures
• But what about the rest of the genome?

4
The advent of multiple genomes

• Since the advent of the mouse genome and others
  there has been an explosion of non-coding
  papers
• detection of ortholog genes
• comparison of up- and down-stream sequences
• test sequences in the laboratory
• publish!
• Some of the shortcomings
• there is no structured annotation of these
  regions
• little clarity on the methods to be used to
  annotate different types of potential regions
• strong focus on mammalian genomes

5
The non-coding world

• There are several features that should be
  detected and annotated in the non-coding world
• Coding features! (I.e. bits and pieces that have
  been missed)
• ncRNAs
• Splice regulators
• Promoter regulatory regions
• Enhancers
• Etc.etc.
• Each of these features deserves its own analysis
  protocol and annotation!

6
Proposed work

• We propose to fill the annotation gap
• Automate large-scale sequence comparisons of
  non-annotated regions
• Focus on all chordate genomes, include Chicken,
  Fugu, Zebrafish, Ciona i., Ciona s.
• Differentiate sequence comparison methodology
  based on evolutionary distance
• Analyze detected regions for potential features
  (e.g. nCRNA, coding, regulatory) and initiate
  appropriate analysis pipeline

7
Gene-centric approach

• Most current large-scale analysis is
  genome-centric, analyzing chunks of sequences at
  a time
• We will focus our analyis workflows on single
  genes of interest, or on single gene families
• Benefits
• for families of particular interests it gives us
  the ability to curate and tailor-cut the analysis
  (e.g. disease genes)
• we will obtain results that are closer to the
  biologists view of the data, with a complete
  analysis of their genes of interest, paralogs,
  orthologs, etc.
• we can tell the biologist how much resemblance
  a gene or gene family bares in a model organism
  genome as compared to the well-studied genomes

8
Workflow design
Coding analysis
ncRNA analysis
Human
Gene of interest
Mouse
Reg. region analysis
Rat
Gene family
Chicken
Synteny Phylogeny
Fugu
Ciona
Conserved Regions
Orthologs and paralogs
ALIGNMENTS
9
Gene family analysis

• Establish ortholog and paralog relationships in
  all analyzed genomes integrating
• Phylogenetic analysis
• Synteny analyisis
• Manual intervention on curated gene families
• Annotate genes for conservation of features
  across genomes
• e.g. show conserved and missing protein domains
  in each genome
• Immediate view of conservation of function for
  biologist

Map Tree to Genome!
10
Alignments

• Analyze sequence surrounding orthologs
• Using 1 Mb on each side of the gene to allow for
  distant enhancers
• Perform global alignments
• MLAGAN alignment, particulary useful in mammalian
  genomes and probably chicken
• Detect conserved regions using varying thresholds
  according to species compared
• Analyze detected regions further using
  promoterwise in more distant genomes
• Allows flipping and rearrangement of short
  regions, likely to happen when comparing human to
  fish

11
Why promoterwise?
Human vs. mouse p63 region VISTA plot
12
Analysis of conserved regions

• Score conserved region
• Base score on conservation in genomes but
  allowing for the fact that it could be missed
  altogether in some genomes
• Analyze for potential features such as
• ncRNAs (gaps in alignments)
• Coding regions (third base mutation bias)
• Regulatory regions (clustering of mutations)

13
ncRNA annotation

• MFOLD structure analysis
• QRNA and DDBRNA analysis
• Compare to RFAM to allocate to known RNA families
• Check for potential antisense activity on
  surrounding genes

14
Coding region annotation

• If identified as coding, define whether
• Portion (exon or part of exon) of surrounding
  gene structure mis-annotated
• Short single-exon gene (should have viable
  promoter, polyA tail,etc)
• Recent pseudogene, relic of evolution

15
Regulatory region annotation

• Perform Transfac search to identify known TFBS
• Analyze all vs. all identified regions to build
  vocabulary of novel potential regulatory regions
• Build free grammar models of regions conserved
  and over-represented in the genome
• With specific gene families characterize regions
  in-vivo, use results to refine models of
  functional regions

16
CODE

• Telethon funded project to perform comparative
  analysis on selected disease gene families
• Integrate work described with manual curation of
  each family
• Keep curated database of features important for
  the disease, e.g. alternative splice variants,
  mutation sites,etc.
• Identify those features automatically in genomes
  analyzed
• Tight collaboration with TIGEM groups
• verifying in vivo our in silico results
• Add experimental info to the database, such as
  in-situ results, cellular sublocalization
  results, yeast-2-hybrid assays,etc.

17
TransCODE

• FP6 (hopefully!) funded project on
  characterization of transcription factors
• Integrate in-vitro SELEX-SAGE assays
• Run all Ciona TFs (600) and some vertebrate TFs
  on column, wash with all degenerate x-mers
• Sequence all x-mers that binding affinity to TF
• Build refined models of TF binding site
• Compare model between Ciona and verterbate
  (hypothesis being that they are very similar!)
• Test identified regions in mouse, zebrafish
  (transient expression assay), Ciona
• In Ciona and zebrafish verify direct action of a
  TF by co-injecting reporter construct with
  morpholino of hypothesised TF

18
Acknowledgments

• Telethon CODE and CORE grants!
• Bioinformatics group at TIGEM
• Pedro Cruz (ncRNA analysis)
• Vincenza Maselli (gene family analysis)
• Remo Sanges (Alignment pipelines)
• IT core
• Giampiero Lago (Webmaster)
• Marco Savarese (Helpdesk)
• Mario Traditi (System Administrator)