Proposed update to the GO wiki IRD page: draft
Draft text for Inferred from Rapid Divergence (IRD),
last reviewed January 30, 2018. It has not been posted. The behaviour it describes was
measured on PAINT's current IBD.gaf and leaf GAF (see ../IRD_EVIDENCE.md);
the PAINT team should confirm it before the page changes.
Questions for the PAINT team first
- Are gene-level IRD rows, like the page's TAIR CPuORF7 example, still produced anywhere?
None appear in the current PAINT exports: the leaf GAF has no IRD evidence and no
IRD-only NOT|IBA rows. - Is "an IRD stops the ancestral IBD from reaching the genes below, and exports nothing
for them" the intended behaviour, or should an IRD also emit NOT|IBA rows as IKR does? - Should an IRD whose ancestral node no longer carries the IBD (14 such rows today) be
removed when the IBD is withdrawn?
Proposed text
Overview
IRD (Inferred from Rapid Divergence, ECO:0000321) is a PAINT evidence code. A PAINT curator
uses it to record that a function annotated by IBD at an ancestral node has been lost, or
diverged beyond recognition, in one clade of the family tree. The IRD is a NOT annotation
on the internal node where the change occurred.
An IRD blocks propagation: the IBD's term is not inherited as IBA by the genes below the
IRD node. It does not create NOT annotations on those genes, so a gene below an IRD simply
receives no IBA for the term. This differs from IKR (Inferred from Key Residues), which
records a loss inferred from missing key residues and does propagate, as NOT|IBA rows on
the genes below.
Use of the With/From field for IRD
With/From holds the PANTHER node (PANTHER:PTN...) of the ancestral IBD whose term the IRD
stops. The reference is GO_REF:0000033.
Example
The ancestral node PTN002538135 (PTHR21646) carries an IBD for cysteine-type deubiquitinase
activity, seeded by experimentally characterised USPs:
| Node | Qualifier | GO ID | Reference | Evidence | With/From (abridged) | Date |
|---|---|---|---|---|---|---|
| PTN002538135 | GO:0004843 | GO_REF:0000033 | IBD | UniProtKB:Q8NFA0, SGD:S000002476, ... | 20260529 | |
| PTN001922764 | NOT | GO:0004843 | GO_REF:0000033 | IRD | PANTHER:PTN002538135 | 20260529 |
PTN001922764 is the eukaryote-wide USP39/Sad1 clade (121 members). USP39 lacks the
conserved catalytic Cys and His of USP-type deubiquitinases and is a tri-snRNP component,
so the IRD stops the deubiquitinase IBA from reaching human USP39 and its orthologs. In
the PAINT leaf GAF, human USP39 (Q53GS9) has no row for the term: no IBA and no NOT.
For contrast, an IKR on the vertebrate PGM5 clade (PTN002731341, NOT phosphoglucomutase
activity) does reach the genes, for example:
| DB Object ID | Symbol | Qualifier | GO ID | Evidence | With/From |
|---|---|---|---|---|---|
| UniProtKB:Q15124 | PGM5 | NOT|enables | GO:0004614 | IBA | PANTHER:PTN002731341|PANTHER:PTN000501426 |
When IRD should NOT be used
- When residues required for the function are lost: use IKR, which records the evidence
and propagates NOT|IBA to the genes. - When only some members of the clade have diverged: place the IRD on the smallest node
that contains them, or annotate the divergent genes directly with experimental evidence. - When an experimentally characterised member of the clade still has the function: such a
gene is often a seed of the IBD being blocked. This was a recurring pattern among the
contested IRDs in our survey (for example S. cerevisiae SSA1 under the fungal
protein-refolding IRD in PTHR19375).
Quality control checks
- The node in With/From carries an IBD for the same term (otherwise the IRD blocks nothing).
- No gene below the IRD node has a positive experimental annotation for the term, unless
PAINT re-annotates the term on a lower node that contains it. - The IRD node lies below the IBD node in the family tree.
Remove
The TAIR CPuORF7 example (TAIR:AT2G31280, NOT GO:0004674, With/From
PANTHER:PTN000357291). PTN000357291 is absent from the current IBD.gaf, AT2G31280 now
maps to bHLH155 (Q58G01, PTHR46196), and no current PAINT output contains a gene-level IRD
row.