Skip to main content
Identity keeps values apart by default, and a relationship is what lets two of them join. But not every biological mapping is a clean one-to-one equality. Some drop rows, some approximate, and some relate an interval to the positions inside it. The ontology makes that imperfection explicit on the relationship itself, so a query can see the risk instead of inheriting it silently.

​
is_lossy

A lossy relationship maps to the same entity but does not preserve every row — a sequence maps to a structure that only covers part of it, or a mapping that exists only where an alignment was found. The join is still type-safe; it just warns that absence of a match is not evidence of absence.
Under strict_joins, lossy edges still join, but the recipe author has declared, in the ontology, that partial coverage is expected here.

​
is_range_mapping

A range mapping relates a position to an interval that contains it, rather than matching two equal values. SIFTS observed ranges, genome-interval overlaps, and flanks all work this way: the join condition is containment, not equality.
Marking the relationship as a range mapping is what stops the compiler from treating start/end as ordinary equality keys — and what tells the caller the result is an interval relationship.

​
coordinate_notes

When a mapping is exact but carries a caveat, the relationship spells it out:
structure_residue_number equals residue_features.position exactly — but confirm wildtype_matches before trusting the mapping.
These notes travel with the ontology, so the warning is visible at query-planning time rather than discovered in the data.
Declared, never inferredThe ontology never guesses that two identities are the same, and it never hides that a mapping is imperfect. is_lossy, is_range_mapping, and coordinate_notes are declarations — the deliberate cost that buys a query the ability to reason about its own uncertainty.
Next: Relations — the typed sources these relationships connect.