1. The Consensus CoDing Sequence (CCDS) Database
Kim D. Pruitt
Mouse Genome Annotation Summit Meeting
March 12-13, 2008

2. Why is the CCDS project needed?
The Problem:
Annotation of the genome sequence is essential
– but beware of different interpretations!
The availability of the human and mouse genome sequence has had a significant impact on disease and health research.
Most scientists rely on annotation information when designing, interpreting, and evaluating research results.
Inconsistencies in annotation results among the main public resources hampers use of this important data.
Researchers may not realize that a different annotation result is available elsewhere – possibly leading to erroneous or incomplete interpretations.

3. CCDS - A collaborative project
Initiated by the main public annotation/browser groups to address concerns by the scientific community about inconsistencies in the human and mouse genome annotation.
Built by consensus among the collaborating members, which include:
European Bioinformatics Institute (EBI)
National Center for Biotechnology Information (NCBI)
University of California, Santa Cruz (UCSC)
Sanger Institute (WTSI)

4. What is the CCDS project?
Project Goals
identify a core set of protein-coding genes that are consistently annotated and of high quality
support convergence toward a standard set of gene annotations
Scope:
Human and mouse protein coding regions
Update frequency
Variable
Depends on frequency of genome annotation updates

5. Process flow – calculating updates
NCBI (computational)
Havana (manual)
Ensembl (computational)
RefSeq (manual)
Compare
CDS
(Annotation +
Sequence)
Ensembl
merged
annotation QA
Identical
Similar
Novel
Existing CCDS
Retain
Lost
New match
New CCDS ID
Out of scope

6. Quality assessment tests include:
Assessing Quality
CCDS status is conservatively applied:
Annotated CDS coordinates are identical
Annotation is of high quality and passes QA tests, or curator review
Existing CCDS proteins can be flagged for review by the collaborating members
Updates and removals are by consensus agreement.
Quality assessment tests include:
Consensus splice sites ("GY..AG" or "AT..AC")
Valid start and stop codons with no internal stops
NMD
Low complexity
Repeat-containing
Insufficient protein homology
Genome conservation
Putative pseudogene
QA test results are reviewed by curators
Over-rides are set to retain supported CDSs

7. CCDS Counts
Date
Build
CDS IDs
GeneIDs
Mar-05
Hs35.1
14,795
13,142
Feb-07
Hs36.2
18,290
16,008
Oct-06
Mm36.1
13,374
13,014
Nov-07
Mm37.1
17,707
16,893
Step Source Genes Proteins
Annotation NCBI
Annotation Ensembl
Matching CDS
QA & curation rejections
Accepted rejections
Final CCDS ID

8. Curation – how are updates curated and coordinated?
Any member of the collaboration can flag a CCDS for review
Update the CDS definition (alter N-terminus extent internal splice site)
Withdraw the CCDS ID (insufficiently supported, or non-protein coding)
NCBI provides a collaboration web site to coordinate this review
All collaborators must agree with a change to finalize a decision
Withdrawal of a CCDS may happen between genome annotation updates
An update to a CCDS is indicated by:
Status change: a status of ‘pending update’ is reported when there is collaborative agreement that a change is needed
Version change: The CCDS version number is incremented once the change is reflected in public annotation. This only occurs after a genome annotation update and CCDS analysis has taken place.
CCDS curation is fully integrated with RefSeq curation

9. CCDS update & curation stats
Curation-based changes:
Mouse: ~5200 curated CCDS genes
name action status count
human update pending
human update agreed
human withdraw pending
human withdraw agreed
mouse update pending
mouse update agreed
mouse withdraw pending
mouse withdraw agreed
923
709
242 24
Annotation pipeline-based changes:
name build status count
human 35.1 Withdrawn, inconsistent annotation
human 36.2 Withdrawn, inconsistent annotation
mouse 36.1 Withdrawn, inconsistent annotation
mouse 37.1 Withdrawn, inconsistent annotation

10. Curation considerations
Alignments
Track low quality sequences (‘kill list’)
Protein conservation
Publications
Personal communications
QA measures

11. Access – How do I know if an annotation has a CCDS ID?
Genome browser displays
NCBI
UCSC
Gene reports
Ensembl
Vega
Other:
RefSeq annotation (NCBI)
CCDS web site
FTP

12. NCBI Map Viewer (chr.5)
Link to
CCDS Browser

13. UCSC Browser
chr5:

14. UCSC Browser – Tyms gene
CCDS Browser

15. Access of CCDS data at NCBI
CCDS Database & Browser interface
Project Description
Query support
Reports attributes of the CCDS
Location data
Sequence members
Status
FTP reports

16. CCDS Browser History Find all CCDSs for the Gene Entrez Gene
View CCDS Details

17. CCDS Browser Mouse-over highlights codon
Click to highlight codon and corresponding amino acid

18. Biology is complex – some CCDS curation examples
1 vs 2 vs ‘n’ genes
translation start site

19. 1 vs. 2 vs. ‘n’ genes Curation Considerations: Nomenclature
History (scientific use, publications, etc.)
Different (but similar) products vs. distinct products
Shared promoters

20. carnitine palmitoyltransferase 1b,
choline kinase beta

21. Current RefSeq representation of the region - two protein coding loci
1 vs. 2 vs. ‘n’ genes
Current RefSeq representation of the region
- two protein coding loci
- one non-coding loci for the non-coding transcript product (a read-through transcript)
Chkb (CCDS )
Cpt1b (CCDS )
Chkb-cpt1b (PMID: )

22. Translation start site
Curation Considerations
Publication reports (CDS begins at ‘n’)
Other cDNA sequencing reveals the ORF can be extended further upstream
Evaluate:
Genome conservation
Literature reports for the protein
Putative Kozak signals
Presence of in-frame upstream stop codon
INSDC submissions from an experimental lab source that do have the longer ORF extent annotated.
Consult with an expert

23. Internal CCDS browser (restricted access)
Jmjd2d jumonji domain containing 2D (chr 19)

24. Update is agreed on by all parties
Resulting in a
258 aa N-terminal
extension

25. Examples – no CCDS ID
EBI+WTSI and NCBI transcript annotation may differ even though the gene includes annotations with CCDS IDs

26. Examples –no CCDS ID Reasons: not found by one group
different CDS length
different splice sites
different internal exon
Curation removal
EBI/WTSI
NCBI
EBI/WTSI
NCBI
EBI/WTSI
NCBI
EBI/WTSI
NCBI

27. Acknowledgements Collaborators at Ensembl, UCSC, Vega Donna Maglott
Josh Cherry
Keith Oxenride
Craig Wallin
Andrei Shkeda
RefSeq Curators
NCBI Genome Annotation Group
NCBI Map Viewer Group
Collaborators at Ensembl, UCSC, Vega
Jen Ashurst & Vega curator group
Rachel Harte
Mark Diekhans
Steve Searle

28. Ensembl – Tyms gene

29. Vega browser Tyms gene (chromosome 5 30388989-30404404)