Manually curated and computationally predicted GO annotations at

Manually curated and computationally predicted GO annotations at the Saccharomyces Genome Database http: //www. yeastgenome. org/ Eurie L. Hong, Ph. D. Department of Genetics • Stanford University School of Medicine

Scientific community Integrated data Analysis tools Data from traditional experiments Data from high through-put experiments Genome sequence

CHS 6/YJL 099 W Locus Summary Page Nomenclature Summary of published data Curated data from� published literature Sequence Information Links to other databases Links to SGD tools and other databases Data from high throughput experiments

Accessing the data via files ftp: //ftp. yeastgenome. org/yeast/

Display of GO Annotations

Status of GO Annotations at SGD All protein and RNA gene products have been annotated with GO terms All GO annotations are manually curated from literature (no IEA) Genes without published characterization data Molecular Function 2112 genes (33. 6% of all genes) Biological Process 1448 genes (23. 0% of all genes) Cellular Component 864 genes (13. 7% of all genes) from Genome Snapshot 8/23/2006

Sources of Computationally Predicted GO Annotations 1. Inter. Pro domain matches in S. cerevisiae proteins source: GOA project 2. Integrated analysis of multiple datasets source: publications, external databases

CHS 6/YJL 099 W Locus Summary Page

Identifying Types of GO Annotations

CHS 6/YJL 099 W GO Annotation Page Core GO Annotations from Large Scale Experiments Computationally Predicted GO Annotations { { {

Changes to GO Term Finder Current functionality { { Specify background set { Refine annotations used by annotation source or evidence codes

Improving GO Annotations Computationally predicted GO annotations Manually curated GO annotations 1. Computational predictions may indicate publications that were overlooked 2. Review inconsistencies between computationally predicted and manually curated GO annotations to improve mappings and manually curated annotations 3. Review inconsistencies between computationally predicted and manually curated GO annotations to improve ontology

Additional Annotations Using Interpro 2 GO Information added to genes with no published characterization data Molecular Function 468 genes Biological Process 316 genes Cellular Component 207 genes from gene_association. goa_uniprot 7/2006

Preliminary Comparison: Cellular Component Annotations Interpro 2 go annotation is ancestor of curated annotation 43% Shared parent is root term 2% Other 38% Interpro 2 go annotation for an unknown 4% Shared parent is child of root term 18% Interpro 2 go annotation matches curated annotation 15% 5946 IEA 9059 IC+IDA+IEP+IGI+IMP+IPI+ISS+NAS+RCA+TAS Other shared parent term 18%

Summary 1. Currently, all GO annotations for S. cerevisiae gene products are manually curated from literature 2. SGD will incorporate computationally predicted GO annotations that will provide additional information for a gene product’s role in biology 3. Computationally predicted GO annotations will be used to refine and improve manually curated GO annotations at SGD

yeast-curator@genome. stanford. edu