Using IDA and has_input for protein binding annotations

[vanaukenk]

Re-visit the idea of annotating protein binding using the IDA evidence code and the annotation extension has_input wrt better aligning conventional and LEGO annotations.

Points to consider:

1. Not all groups use annotation extensions
2. IPI annotations used as input for other tools, e.g. Cytoscape
3. Enforcing appropriate cardinality in annotation extensions for terms like protein binding, bridging where the bridging molecule may bind 2 or more other molecules
4. Examples of how to handle multi-species interactions

[RLovering]

@RLovering interested in joining any further discussions on this

[ValWood]

Me too. We do this conversion internally, http://www.pombase.org/spombe/result/SPBC11B10.09 but export to GO as "with" field Although we use the display label "binds" rather than has substrate.

Thoughts on the above considerations: For

3. Enforcing appropriate cardinality in annotation extensions for terms like protein binding, bridging where the bridging molecule may bind 2 or more other molecules

Curation tools can easily prompt for 2 binding partners for "protein binding, bridging" type terms (Canto already does).

For

2. IPI annotations used as input for other tools, e.g. Cytoscape Cytoscape are missing a trick. Extensions are already a rich source of network data that they are presumably ignoring (protein kinase x has_substrate y). Most of PomBase network data comes from these extensions now 2058 from substrates vs 1420 from binary interactions:

see: http://www.esyn.org//builder.php?type=Graph&term=GO:1901990&interactionType=physical&source=pombase&includeInteractors=false (all orange edges with little arrow heads are effector-substrate pairs)

So using extension field for IPI instead of "with" and pointing users to the richer source in annotation extensions would benefit them a lot.

[ukemi]

I will keep this tagged as a GOC meeting topic and we can put it on the agenda for Oregon.