InterPro Mapping Review Project

IN_PROGRESS PIPELINE

Collection: Propagation by homology · Function prediction evaluation

Warnings (3)

InterPro Mapping Review Project

Bottom line: InterPro2GO (GO_REF:0000002) attaches GO terms to every protein
that matches an InterPro entry, so a term that holds for only some members of a
family is copied onto all of them. We harvested every InterPro2GO annotation our
gene reviews had already judged (3,652 records on 1,706 genes, joined to 1,826
source entries), ranked the entries by how often reviewers flagged them, and ran
family-level deep research on the top of that list. The result so far is a set of
36 proposed mapping edits across fifteen entries: ten removals (for example ATP
binding and protein phosphorylation on the protein kinase domain, IPR000719), twelve
proposals to narrow or qualify a mapping, and fourteen endorsements. These are
proposals for InterPro curators, not changes InterPro has adopted, and the
worklist below the first dozen entries is not yet assessed.

We did this because a wrong InterPro2GO mapping is a single error that repeats
across every matched gene, so one fix at the mapping level is worth many
gene-level corrections. Gene reviews also catch mappings the ranked worklist
never reaches: three of the fifteen entries came from reading human genes one at
a time, and three more from auditing the bacterial export ATPases FliI and SctN
(see conserved-function recovery).

Start here: Proposed mapping edits ·
Prioritized entry worklist ·
Gene-level evidence and reasons

Findings and proposed curation actions

The mapping set records the individual GO terms,
verdicts, and rationales. The following summarizes its recommendations; family-level
research proposals still require checking against the entry's membership and
experimental evidence.

InterPro entry Finding in the family assessment Proposed action
IPR000719 — Protein kinase domain Pseudokinase members challenge the assumption that every domain match binds ATP and phosphorylates proteins. Remove the blanket ATP-binding and protein-phosphorylation mappings; assess narrower catalytic entries.
IPR001128 — Cytochrome P450 The assessment supports cofactor-binding mappings but questions uniform catalytic assignments across functionally diverse members. Retain the binding mappings; review the scope of monooxygenase and oxidoreductase mappings.
IPR001424 — Cu/Zn superoxide dismutase domain Copper-chaperone members make the domain-wide superoxide-metabolism assignment too broad, while fungal Cu-only SODs show the term still needs cover outside the canonical Cu/Zn SOD rule. Remove the blanket process mapping and request process-term cover on both the Cu/Zn SOD UniRule and a fungal Cu-only SOD rule or narrower entry; assess metal binding separately.
IPR000276 — Rhodopsin-like GPCRs The assessment flags atypical receptors as exceptions to canonical G-protein coupling. Review receptor-activity and signaling mappings at subfamily level, verifying evidence for the proposed exceptions.
IPR001046 — NRAMP / SLC11 The assessment supports broad metal-transport terms rather than substrate-specific assignments across the entry. Retain broad transport mappings; evaluate more specific functions on individual members.
IPR012724 — Chaperone DnaJ The assessment attributes ATP binding to the Hsp70 partner rather than DnaJ. Remove the ATP-binding mapping; retain protein folding and review the scope of the heat-response mapping.
IPR007197 — Radical SAM Diverse reactions can justify a broad family-level molecular-function term. Retain catalytic activity and iron-sulfur-cluster binding; avoid unsupported specialization across the domain.
IPR020849 — Ras-type small GTPases The assessment proposes GTPase activity in addition to GTP binding. Check the proposed addition against family-wide catalytic competence; narrow process and localization assignments where needed.
IPR002100 — MADS-box domain The assessment distinguishes domain-level DNA binding and dimerization from whole-protein transcription-factor function. Retain the two domain-level mappings; do not infer transcription-factor activity from the domain alone.
IPR045122 — Calcium permeable stress-gated cation channel 1-like The entry name says calcium permeable, but GO:0005227 is defined as a channel that opens when calcium binds it — opposite directions of causation. The TMEM63/OSCA members are stretch- and osmolarity-gated. Remove the calcium-activated mapping; assess mechanosensitive cation-channel and calcium-channel terms in its place.
IPR042371 — Z-binding domain A catalytic activity is mapped onto a binding domain: the deaminase belongs to ADAR1's separate domain, so ZBP1 inherits an activity it has no domain for. Remove the deaminase mapping and restrict it to entries carrying the catalytic domain; retain RNA binding.
IPR013380 — Type 3 secretion system ATPase SctN An SctN-specific entry mapped to rotational proton-transporting ATPase activity and ATP biosynthesis. SctN is a soluble protein-exporting ATPase with no Fo partner; the terms are wrong for every member (915 annotations). Remove GO:0046961 and GO:0006754; retain the type III secretion process and complex mappings; consider adding GO:0008564 protein-exporting ATPase activity.
IPR004100 — ATPase, F1/V1/A1 α/β, N-terminal domain Sound for rotary ATPase subunits, but the domain is shared by the FliI/SctN export ATPases, which do not transport protons. Narrow GO:1902600 and GO:0046034: ideally suppress when IPR005714 (FliI/YscN) also matches; if per-entry mappings cannot express that, remove them or move them to narrower entries that exclude FliI/SctN.
IPR005714 — ATPase, type III secretion system, FliI/YscN GO:0009058 biosynthetic process remains in the 2025-09 mapping file, a remnant of the ATP synthase ancestry. Remove it; the entry's other four mappings are sound.
IPR006935 — Helicase/UvrB, N-terminal The entry is dominated by DNA-acting enzymes, but the same fold occurs in the RNA-sensing RIG-I-like receptors, so IFIH1 (MDA5) inherits DNA binding. Replace the DNA-binding mapping with GO:0003676 nucleic acid binding, as on the sibling helicase entries IPR011545 and IPR006555; the nucleic-acid substrate is a property of the enzyme, not of the shared fold. ATP binding and hydrolase activity are unaffected.

Rule cover for the removals

Each removal row in the mapping set ends with a note on what still supplies the term to the members
that should keep it. Where a term is right for only part of an entry, a conditional UniProt rule is
often the better carrier than a mapping, because a rule can require a site, exclude a domain, or limit
the taxon. The notes were checked against the UniProt UniRule API and the current interpro2go file on
2026-10-10.

Removal What still supplies the term Request
IPR000719 ATP binding IPR017441 (ATP binding site) for members with the motif; family rules such as the JNK rule UR001528758 for members without it Remove; all 7 reviewed sole-source losses were accepted or kept, so family rules are needed
IPR000719 protein phosphorylation Active-site entries IPR008271 and IPR008266 Remove
IPR001424 superoxide metabolic process UniRule UR000000113 selects canonical Cu/Zn SODs and excludes CCS, but assigns only the EC number; Candida albicans SOD5 lacks its PS00332 condition and needs separate Cu-only SOD cover Remove; add GO:0019430 to the Cu/Zn rule and to a fungal Cu-only SOD rule or narrower entry
IPR012724 ATP binding Nothing needed; the HAMAP DnaJ rule MF_01152 also assigns no ATP-binding or zinc-ion-binding GO term Remove
IPR042371 dsRNA adenosine deaminase activity IPR002466 gives ADARs a deaminase term; no rule assigns the specific one Remove; a rule conditioned on IPR002466 could restore it
IPR006935 DNA binding HAMAP UvrB rule MF_00204; nothing for RIG-I-like receptors or Dicer Replace with GO:0003676
IPR045122, IPR013380 (two), IPR005714 Not needed; the terms are wrong for every member Remove; for IPR045122 and IPR013380, add the correct family-wide terms as mappings

The checks themselves are described in Before proposing a removal.

A recurring curation question is whether a function holds across the matched
entry, not simply whether it is well established for one member. Conversely, a
broad term can be appropriate for a diverse family even when a more specific term
is preferable for an individual gene.

Archaeal additions (2026-10-15). Gene reviews in the ARCHAEA
project added six rows to the mapping set, not counted in the totals above:
removals for IPR019982 to threonine-type endopeptidase activity (non-catalytic
archaeal proteasome alpha subunits), IPR018229 to ion channel activity and
IPR017402 to light-activated ion channel activity (rhodopsin pumps); a proposed
Eukaryota restriction for IPR000426 to ubiquitin-dependent protein catabolic
process and IPR005024 to vacuolar transport; and narrowing IPR000534 to NAD
binding. interpro2go has no conditions, so the two restriction rows each name a
curated replacement rule in rules/curated (CR-0001,
CR-0003) that couples the InterPro condition with a taxon or module condition.

What curators can act on

  1. Assess the four removal proposals first: ATP binding (GO:0005524) and protein
    phosphorylation (GO:0006468) for IPR000719, superoxide metabolic process
    (GO:0006801) for IPR001424, and ATP binding (GO:0005524) for IPR012724. Verify the cited exceptions and the current entry scope.
    Each removal row ends with a rule-cover note; check it before forwarding the removal.
    The IPR013380 removals (GO:0046961, GO:0006754) are simpler: the entry is SctN-specific, so
    they are wrong for every member rather than for an exception subset.
  2. Resolve the proposed GTPase-activity (GO:0003924) addition for IPR020849. Confirm that the
    assignment is supported across the entry, including divergent members.
  3. Separate gene-level refinement from mapping-level error. A MODIFY or
    KEEP_AS_NON_CORE decision on one gene does not by itself justify changing a
    family-wide mapping. Record whether a proposal requires removal, a narrower entry,
    or only a more specific annotation on that gene.
  4. Check apparently accepted annotations on exception members. Prioritize
    pseudokinases, copper chaperones, and other members whose functions diverge from
    the mapped activity; update gene reviews when the evidence supports a correction.
  5. Extend the family assessments using the ranked worklist. Each recommendation
    should identify the affected entry and GO term, supporting evidence, and a specific
    requested change for InterPro curators.

Evidence snapshot

The committed annotation table contains
3,652 InterPro2GO annotation records. Its priority table covers 1,826 resolved
InterPro entries, plus an unresolved-source group. These are stored extraction
results, not live counts of the growing review collection.

Gene-review action Count
ACCEPT 1,870
MODIFY 609
KEEP_AS_NON_CORE 523
MARK_AS_OVER_ANNOTATED 391
REMOVE 194
NEW 30
UNDECIDED 15
PENDING 20

The worklist flags 1,732 records (47%): MODIFY, KEEP_AS_NON_CORE,
MARK_AS_OVER_ANNOTATED, REMOVE, or UNDECIDED. This is a review-priority
measure, not an estimated mapping error rate. It includes valid but non-core
annotations, specificity refinements, and unresolved judgments. NEW and PENDING
records are excluded from this flag.

Entries with repeated gene-level flags

These counts prioritize investigation; they are not family-level verdicts. A record
can cite several entries, so counts across entries should not be summed.

InterPro entry Annotation records Flagged records
IPR000719 112 50
IPR008271 55 34
IPR001128 44 26
IPR036396 44 26
IPR001424 20 16
IPR036423 20 16
IPR001046 27 15
IPR012724 24 13
IPR000276 21 12

See the complete worklist for action
breakdowns, example genes, and affected GO terms.

Supporting material