Browse
Executable evidence boundary

Comparison laboratory

The server evaluates identity, endpoint, value, assay, condition, molecular-form, dependence, direction, evidence-class, and provenance rules. The interface cannot override a refusal.

Corpus records

Select two admitted measurements.

Authoritative decision

Comparison refused

Construct mismatch · Canonical full-length and truncated constructs cannot be ranked as one target state.

LEFT / fixture-construct-aP00918 · Ki

= 12 nM · biochemical · pH 7.4

RIGHT / fixture-construct-bP00918 · Ki

= 12 nM · biochemical · pH 7.4

VARIANT_OR_CONSTRUCT_MISMATCH

variant, construct

The variant or experimental construct differs.

Needed: Match sequence variant, truncation, fusion and tag state.
Policy principles

Refusal is a first-class result.

  1. Compare only exact, versioned protein identities; names are not identifiers.
  2. Rank quantitative measurements only within one assay or an independently reviewed matched-protocol group.
  3. Keep Ki, Kd, IC50, EC50, Km, kcat and kcat/Km as separate endpoint types.
  4. Preserve source relation operators and raw values; a bound is not a point estimate.
  5. Keep experimental constructs, variants, tested molecular forms and conditions explicit.
  6. Display a refusal with actionable reasons whenever strict comparison requirements are not met.
Download the machine-readable policy →