Every KG needs primary knowledge source for every edge

[capasfield]

By March 24 in DEV:

Every KP edge MUST return a biolink:primary_knowledge_source attribute. All of the attribute values MUST come from the infores repository: https://github.com/biolink/biolink-model/blob/master/infores_catalog_nodes.tsv

If a source does not exist in the catalog, please make a PR adding it to the catalog.

[mbrush]

I might add to this rule/convention the notion that KPs that generate new knowledge by analyzing data SHOULD report the sources of the data that they operate on - using the biolink:supporting_data_source property. The UI team has indicated that this is important information for end users to see.

Some examples of this:

• EHR or environmental data sources used by ICEES or COHD to generate variable correlation edges
• Clinical/EHR and omics (e.g GTEx, TCGA, GDSC) data sources used by Multiomics KP to generate clinical variable, gene expression, and other correlations
• TCGA as a source of data used by CHP to generate various types of survival time edges
• CMAP and CTRP and DepMap as sources of data used by MolePro to generate chemical and gene similarity edges
• LINCS as a source of cell line expression data used by Improving Agent to create various gene correlation/upregulation edges
• Pubmed database as a source of textual data used by TMKP to generate text mined associations
• Medline database as a source of textual data used by Improving Agent phenotype/co-occurrence edges

Note that I added a box to the top of the TRAPI Retrieval Provenance Standard document outlining the key rules and conventions, including this one. See here.