Copy of biocuration 2017
1 like215 views
This document discusses using co-annotation analysis and biological knowledge as a quality control procedure for gene ontology (GO) annotations. It describes a multi-step iterative process where GO term intersections are identified and used to generate rules of expected relationships. Violations of these rules are reported to contributing databases and used to identify annotation errors. The process led to over 100 rules being created and 83 rule violations found by analyzing 568 gene annotations across several species. Future work involves implementing the rules directly into the GO pipeline and expanding the approach to additional biological processes and functions.
Copy of biocuration 2017
- 1. Using co-annotation and biological knowledge as a quality control procedure for ontology structure and gene annotation in the Gene Ontology Biocuration, 2017 Valerie Wood and Seth Carbon @news4go
- 2. No co-annotated genes (e.g.) • Unlikely there is a protein that is directly involved in both processes. • Use this biological knowledge to check annotations. tRNA metabolism transmembrane transport 161 339
- 3. Intersections in a simple eukaryote
- 4. Intersections in a simple eukaryote
- 5. The Matrix Tool http://amigo.geneontology.org/matrix * * http://tomodachi.berkeleybop.org/matrix
- 6. Step 1 Annotations shared between sets of GO terms are explored and annotation intersections are noted.
- 7. Step 3 Identify new annotations violating existing rules. Report to contributing database(s) for validation. Step 2 Rules are created for “zero intersects” based on known biology: • (“cellular amino acid meta. proc.” ∩ “DNA recombination”) = 0 • (“lipid meta. proc.” ∩ “carbohydrate meta. proc.”) = 0
- 8. Step 4 Annotations critically inspected, leading to one of two outcomes: A: Violation identified: contributing database corrects annotation B: Annotation confirmed: rules are extended to allow specific intersections: (“amino acid metabolism” ∩ “cofactor metabolism”) = 0, except if 'de novo' NAD biosynthetic process from aspartate (“amino acid metabolism” ∩ “peroxisome organization”) = 0, except if C. elegans prx-10 (peroxisome protein and tryptophan synthase)
- 9. Steps 1-4: an iterative process Explore co-annotation Correct or modify Identify and report Biological “rules”
- 10. Example errors membrane organization 167 1011 INDIRECT EFFECT Brr6 is involved in nuclear envelope organization, when mutated this causes nucleocytoplasmic transport defects, but it is not involved in this process InterPro2GO MAPPING ERROR Brr6 annotation used in electronic mappings (IEA) applied to 814 entries DOWNSTREAM EFFECT Fes1 is a ribosome associated protein involved in clearing misfolded proteins, when mutated can affect translation (upstream process) nucleocytoplasmic transport cytoplasmic translation 450 1823 protein catabolic process
- 11. Fission yeast intersections 01/2012
- 12. Fission yeast intersections 03/2017
- 13. 29 annotation errors corrected Multispecies exercise: cohesin complex vs. processes
- 14. Multispecies rule building • Rule base needs to apply to all species • Used fission yeast, budding yeast, worm, and mouse for phylogenetic coverage and a large body of experimental annotation • Critically evaluated annotation outliers in intersections with slim terms for: amino acid metabolism, tRNA metabolism, translation, ribosome biogenesis, and DNA replication • Fixed errors • Built rules for “zero” intersections • Extended some intersections to allow exceptions
- 15. Multispecies rule building report: http://bit.ly/go-report-int ; data: https://github.com/geneontology/shared-annotation-check
- 16. Multispecies rule building report: http://bit.ly/go-report-int ; data: https://github.com/geneontology/shared-annotation-check
- 17. Multispecies rule building report: http://bit.ly/go-report-int ; data: https://github.com/geneontology/shared-annotation-check
- 18. Multispecies rule building results 107 rules created, 83 rules broken, by 568 experimental annotation violations Types of errors identified (and corrected): 147 74 73
- 19. Multispecies rule building results
- 20. The Future • Implement rules as part of GO pipeline (Jenkins); reports to MODs (Currently reporting at: http://bit.ly/go-report-int) • A community resource; set up a mechanism to challenge rules (GitHub for now: https://github.com/geneontology/shared-annotation-check) • Full “slim” process coverage • Component/complex/function intersections • Regulation vs. “involved in” • Something to make the function prediction people happy
- 21. Thank yous and contact • Chris Mungall • Midori Harris, Antonia Lock, David Hill, Stacia Engel, and Kimberly Van Auken • The InterPro Team • The UniProt Team • Peter D'Eustachio and Reactome • Monica Munoz-Torres and Suzanna Lewis • 104967/Z/14/Z (Wellcome Trust) NHGRI U41HG 002273 (Gene Ontology) • Valerie Wood http://pombase.org vw253@cam.ac.uk @ValWood (GitHub) • Seth Carbon http://berkeleybop.org sjcarbon@lbl.gov @kltm (GitHub)