Biolink.

A graph is only as useful as the agreement about what its nodes and edges mean.

Connect enough biomedical databases and you hit a wall that is not about data, it is about grammar. One resource calls something a "gene," another a "gene product." One says "causes," another "associated with," and they do not mean the same thing. Without a shared vocabulary for the categories and the relationships, you cannot merge two sources without quietly corrupting both.

The Biolink Model, from the Monarch Initiative and the NCATS Biomedical Data Translator, is that shared grammar. It defines the categories (gene, disease, phenotypic feature, variant), the predicates that connect them, and the identifier conventions that name them, so that knowledge from many sources can live in one graph and be queried as one thing. It is the schema the modern biomedical knowledge-graph world has largely agreed to speak.

Biolink is why Sniff is a graph and not a pile of tables.

Every entity in Sniff, a gene, a disease, a variant, a breed, a phenotype, and every relationship between them is typed and identified in the Biolink idiom. That is what lets OMIA, VBO, Mondo, uPheno, and Ensembl coexist in one coherent graph instead of a stack of incompatible spreadsheets, and it is what makes the graph queryable by a tool or an agent without a custom key for every source.

It is the least visible of Sniff's sources and one of the most important. Speaking the same grammar as the wider biomedical knowledge-graph world is also what makes Sniff interoperable with it, rather than a private dialect nobody else can read.