[RESEARCH SPIKE] Create relationship links between resources hosted in CCKP, HTAN, and GENIE

[jaybee84]

Per conversation including @jaybee84, @adamjtaylor, @susheel, @xindiguo, and @aclayton555

In the near future we would like to create links between resources hosted in these repos to add to the "Related Resources" section in CCKP. The expected behavior is the following:

User views a spatial transcriptomics dataset from breast cancer on CCKP >> if an ST dataset from breast cancer exists on HTAN or GENIE, those datasets show up as related resources or "Recommended Resources" to the user on the Dataset Details page on CCKP, helping increase discoverability of all cancer datasets across multiple repositories/Portals.

[Bankso]

Extending the Data Catalog concept might help us build this feature.

If we aim to annotate all datasets (or dataset representations, including folders) of interest with the Data Catalog model and store this as a common table, it would be relatively straightforward to generate a knowledge graph from that information. We could go one step further and apply a common set of keywords or other annotations (disease type, tissue type), which would also be integrated into the knowledge graph.

In terms of using this information, one way could be to extend the Portal search co-pilot idea that we are starting to work on for B2AI, by providing the knowledge graph to the agent. This might be able to support the similarity-based recommendation feature described above.

I imagine content for new Related Resources tabs could be retrieved from the knowledge graph, too. Based on my recent experience with the B2AI standards explorer, we might have to consider how Synapse tables could be used to store the knowledge graph info, but I think this would be doable.