RHEA to GO

What reaction-level annotation adds beyond EC numbers, and 132 curated new mappings

AI Gene Review · projects/RHEA · 2026

Bottom line

  • RHEA reaches GO via rhea2go (GO_REF:0000116), but 88% of reactions that also have an EC map to the same GO term ec2go already gives.
  • Its real value: 772 RHEA-only GO terms, 462 reactions whose EC has no ec2go line; meanwhile 36% of enzymatic reactions have no GO target.
  • We curated 132 new mappings backed by reviewed enzymes: 42 new Swiss-Prot annotations. The residual gap is missing GO terms, not missing mappings.

EC masks most of RHEA

Why audit RHEA

  • RHEA is a curated, reaction-level source of enzyme function, used on UniProt catalytic-activity lines.
  • Gene reviews see GO_REF:0000116 rows; we needed to know if they are independent evidence or EC duplicates.
  • Specificity cuts both ways: 679 GO terms absorb >1 reaction (up to 67 ADH reactions → one term), while 112 ECs split across several GO terms.

Reverse gap: an upper bound

High values are generic-parent reactions where curated entries carry specific child terms. A closure-aware re-run is still needed to find true gaps.

132 curated mappings (rhea2go.sssom.yaml)

Class Rows Meaning
exactMatch 110 ready-to-add; most EC-bridge supported (BTD, TPMT, VKORC1L1, PYCR1, RNGTT)
broadMatch 4 only a class term exists (PHYKPL → lyase activity; B3GALNT2)
NoTermFound 18 new GO term needed (e.g. hppE fosfomycin epoxidase)

Every row backed by a reviewed enzyme (RHEA-MAPPING-REVIEWS.md); passes just validate-rhea-mappings. Gain: 42 new reviewed-entry annotations.

Status and next steps

  • ✅ EC-masking and specificity measured; reverse-gap pilot; gap case reviews (PHYKPL, B3GALNT2, SAMD8, SULT6B1)
  • ✅ 132-row SSSOM set validates; EC-bridge reviewed residual gain is 0
  • ⬜ Forward closure-filtered cross-organism scan (needs go-db)
  • ⬜ Closure-aware reverse gap on RHEA:21248 and RHEA:10596
  • ⬜ Batch new-term requests for reactions with no GO target

Read more: projects/RHEA.md · projects/RHEA/rhea2go.sssom.yaml · projects/RHEA/RHEA-ANNOTATION-GAIN.md