So now I'm way at the back of the main conference all because my laptop battery is not what it used to be...
- Annotations are made by different annotation groups
- How do you maintain consistency
- GO aims to maintain consistency
- Standards, best practice, training
- After the annotations how do we check the consistency ?
- Compare annotations (curated orthologs) across species ?
- Measure of annotation consistency: e.g. Mouse, human and rat annotation of PAX8, how do they compare ?
- Collect data, cluster annotations and make comparison
- Use curated ortholog sets from MGI about 15000
- Focus on human and mouse
- Issues with annotation sets: evidence codes...
- Cluster GO annotations, use GO_Slims: High level grouping terms, use structure of go to infer annotation...
- Can infer annotation up to the go slim level
The slide is showing the GO hierarchy, so the idea (I think) is you take low level terms and follow the hierarchy up to the GO Slim terms (or level) to assess the consistency.
- Confusion matrix summary ??
- Some annotation errors, some due to experimental differences...
This kind of method may be relevant outside of bioinformatics ontologies, maybe we should look into checking the consistency of folksonomies ? or normalization of folksonomies ?
Summary
Can do consistency checking using comparative gene sets...
http://www.spatial.maine.edu/~mdolan
It is a little hard to hear the questions, I'm sitting at the back of the room
I wonder if using GO Slims actually biases the annotation consistency checks....
They are not inferring annotations only seeing what does not match, those that do not are sent back to the curators...
