Annotation Conf. Call 2017-07-25
Meeting URL - NOTE CHANGE TO ZOOM
Agenda
Annotation Guidelines for Modified-Protein Binding
- Pascale and Sylvain will present the guidelines for annotating to modified-protein binding terms
GO Slims
- AGR is including a GO slim on the gene pages
- Launch date for AGR 1.0 is October 2017
- Need feedback on GO Slim and ribbon display on gene pages
- Mary Dolan will provide an overview of the methodology used to generate the slim
Alliance JIRA ticket, curators comment here: https://agr-jira.atlassian.net/browse/AGR-470
Signaling Project
- Update on what pathways have been claimed
- Wnt - Giulia, Helen, Kimberly
- Update on progress to date, github tickets
- MAPK - David H.
- Examining gene products annotated to the process with no experimental support for known MFs
- Thinking about how to define the beginning and parts of the pathway
- Ca2+ Signaling
- Fertilization - Penelope
- GPCR - Pascale, Petra
- Some ontology tickets have been addressed
- Needs global annotation review
- Needs GO-CAM model
- Reminder - discussion on specific github tickets, not on project page
- MAPK - David H.
Minutes
- On call: David H., Edith, Emily, Giulia, Harold, Helen, Jim, Mary, Midori, Pascale, Paul T., Penelope, Petra, Rob, Ruth, Sabrina, Sage, Shur-Jen, Stacia, Stan, Sylvain, Tanya, Tom
Annotation Guidelines for Modified-Protein Binding
- Discussion began at USC meeting in November 2016
- Question: How to correctly annotate when binding to a protein is modification-dependent?
- For example, a protein binding ubiquitin vs protein binding that depends on ubiquitin modification
- Solution: create new terms, using the most common types of protein binding that reflect the dependency
- Less frequent modifications would be annotated to the parent term, modification-dependent protein binding
- Comments on term usage have been added to the terms in the ontology
- For example, methylation-dependent protein binding
- Full documentation will be added to the go-annotation github repository
- Sylvain reviewed annotation errors
- For example, targets of post-translational modifications should not be added to the BP term for the modification
- Annotations that may have been affected by these terms are recorded in the github ticket for this issue
- Curators can review these annotations to make sure that we didn't lose information by merging into the parent term, protein binding
GO Slims
- Mary presented slides on construction of AGR GO slim
- Started with experimental annotations for all organisms included in AGR + human
- Select terms based on number of annotations, depth in DAG, information content using an algorithm
- Select terms
- Tended to select terms that had between 200 - 1000 annotations (don't want terms with 30000 annotations)
- Selected terms that are generally 4-7 levels down in the DAG - note that we can always select a parent if a child term is too specific and doesn't have many annotations
- Eliminate parent-child terms
- Then manually analyze the list
- Analyze what terms are in and out
- Analyze what genes are in and out
- It might be worthwhile to include an 'Other' category to capture the genes that don't currently map to a slim
- In AGR, the ribbon display is the main way, right now, that GO annotations will be displayed
- Gene products can fall into more than one bucket of the slim
- Select terms
- Map selected terms onto the graph view of GO terms to see what is/is not included
- Mary also has a view that shows the breakdown of annotations based on contributing group
- Question about terms in F list
- How many reflect the overall function vs part of the mechanism of the function
- For example, metal ion binding - what types of gene products are annotated to these terms? Transporters? Others?
- How many reflect the overall function vs part of the mechanism of the function
- Mary will make slides available here on the wiki
- Relevant github tickets:
- The AGR slim will be available to all as part of the ontology
- What are the plans, if any, for integrating the AGR slim with the generic GO slim?
- Overall workload - not too time-consuming to generate a slim