Determine type and number of inferred annotations in Noctua models

[vanaukenk]

We need to decide if the inferred annotations derived from Noctua models are adding valuable knowledge to our standard GO annotations.

To do this, it would be very helpful to know what GO terms are inferred by the reasoner and how many of each term we get.

[pgaudet]

@balhoff Can you make a query to get this information?

[vanaukenk]

@pgaudet What project should this go in?

[balhoff]

How would you like to define inferred annotations? It's a little complicated since technically every annotation results from an inference process (e.g., even to just infer 'enables' from 'enabled by'). Two of the main situations I would call inferred are:

• GO class "deepening": inferring a type for a node that wasn't asserted
• connecting gene products to GO terms via extended property chains and transitive property (but many of the basic annotations come from an enables/part_of chain)

[kltm]

This is to confirm the scope of https://github.com/geneontology/project-management/issues/53 ?

[pgaudet]

@balhoff

We are interested inferences that create 'duplicate' annotations. Deepening would be one case. The other type of situation is where you have MF1 > directly positively regulates > MF2 and MF 1 is inferred to have GO molecular function regulator activity

(I think this is the 'connecting gene products to GO terms via extended property chains and transitive property' situation)

ie we want to know when a gene gets a different GO term from the term being annotated

[kltm]

Now in Misc. project.

[suzialeksander]

Bumping, @dustine32 and @kltm were thinking of ways this count can be done for the loads in Noctua, inluding the recent SGD load.

[kltm]

See: https://github.com/geneontology/minerva/issues/539