1. VectorBase genome annotation
VectorBase-EBI, European Bioinformatics Institute, Wellcome Trust Genome Campus, Hinxton UK
VectorBase SWG 2006

2. Overview of current annotation system
Assembled genome
Sequencing centre gene predictions
VectorBase gene predictions
Merge into canonical set
Protein analysis
Assembled genome - Changes to the assembly require significant changes to the annotation process - raw computes, transfer of existing annotations/features
Sequencing centre gene set - tendency to overpredict
VectorBase gene set - tendency to underpredict
Display on genome browser
Release to GenBank/EMBL/DDBJ
VectorBase SWG 2006

3. Merging gene sets Gene set #1 Gene set #2
Reduce to single predictions per locus
Compare exon/intron structures
Identical structures
Compatible structures
Different structures
Merge/Split structures
Complex
No Map
Add isoform predictions based on EST/Peptide data
Canonical gene set
VectorBase SWG 2006

4. Data types used for gene prediction/validation
Protein sequences
‘Self’ (i.e. species to be predicted)
Taxonomic splits of UniprotKB
Transcript sequences
mRNAs
ESTs
Evidence of expression
Microarray
SAGE tags
Ditags
MPSS - Massively Parallel Signature Sequencing - immobilize short DNA fragments on millions of beads, then sequence these oligomers in parallel (I.e. fast). Claims to be able to count 1 million mRNAs at a time
MPSS
Proteomics data
Sequence statistics
Coding potential
Splice site prediction
VectorBase SWG 2006

5. VectorBase gene prediction pipeline
Blessed predictions
Manual annotations
Community submissions
(Apollo)
(Genewise, Exonerate, Apollo)
Similarity predictions
Species-specific predictions
Canonical predictions
(Genewise)
(Genewise)
Protein family HMMs
ncRNA predictions
(Genewise)
(Rfam)
Transcript based predictions
Ab initio gene predictions
(Exonerate)
(SNAP)
VectorBase SWG 2006

6. VectorBase curation database pipeline for manual/community annotation
Community annotation (Community representatives)
Manual annotation (Harvard)
Curation
warehouse db
Chado-XML
Chado-XML
Apollo
Chado
Apollo
Community annotation (in collaboration with Harvard)
GFF3
Ensembl
Gene build db
VectorBase SWG 2006

7. Overview of current re-annotation system
Full gene build
Partial Gene build
New gene build
Blessed genes
Species-specific gene prediction
Current gene set
Compare
Merge
Updated gene set
VectorBase SWG 2006

8. Comparing new gene builds with the old one
Use of manual annotation for validation of automated gene build improvements
Simple statistics (CDS length, intron size, CDS matching TE’s)
BRC annotation metrics
Supporting evidence for a gene prediction (citation,expression,orthology)
Attachment of Standard Operating Procedures (SOPs)
VectorBase SWG 2006

9. Gene build schedules Triggers for re-annotation Temporal Data
Full gene build
Triggers for re-annotation
Temporal
Data
New EST data for species
New genomes
Re-annotated genomes
4 months
1 month
Partial gene build
Temporal - It’s been a while (6 monthly, yearly)
Data - New transcript data = better predictions
- New genomes within claid, comparative approaches
- Re-annotated genomes, e.g. new ESTs in Aedes influences prediction changes which affect Anopheles entries, GPCRs in nematodes
VectorBase SWG 2006

10. VectorBase annotation capacity with increased number of genomes
Gene builds per year per genome
2 full 1 full 1full 1 full partial 3 partial 2 partial 1 partial
2 genomes Yes Yes Yes Yes
3 genomes Yes Yes Yes Yes
4 genomes No Yes Yes Yes
5 genomes No Yes Yes Yes
6 genomes No No Yes Yes
7 genomes No No No Yes
8 genomes No No No No
VectorBase SWG 2006

11. Re-annotation questions
Triggers for re-annotation
Strict temporal triggers
Always do a full gene build every year?
Data triggers
How much new data is enough?
Knock-on effects of related species (re)annotation?
Encouraging community submissions
How can we get more community annotation input?
Outreach at conferences (Roadshow)
VectorBase SWG 2006