PANTHER IBA family review

Testing 160 fission yeast IBAs at the tree node they came from

AI Gene Review · projects/PANTHER_IBA_REVIEW · 2026

Bottom line

  • Each IBA follows from a PAINT curator's IBD at an ancestral PANTHER node; we rebuilt that propagation for all 160 IBAs on 41 reviewed S. pombe genes.
  • The per-gene calls held up: 148 kept, the 36 cross-subfamily flags were mostly conserved functions, no new errors among accepted rows.
  • Real over-propagations were localization terms and paralog-specific functions crossing subfamily lines (pom1, rqh1, mid1).

What an IBA asserts

  • A PAINT curator reads the tree, the MSA and the experimental annotations, and places the function at a node (IBD). Descendants inherit it as IBA.
  • So an IBA is a phylogenetic judgment. To challenge one, ask: is the target inside the clade that inherited the function, and is there target-specific evidence of loss or divergence?
  • A short seed list is not weak support; one well-characterized descendant can ground a node.

Method: rebuild each propagation from repo data

  • with/from → source node (PTN…) and seed genes.
  • Gene and seed subfamilies from *-uniprot.txt and interpro/panther/<FAM>/<FAM>-entries.csv.
  • Node-level PAINT annotations and IRD/IKR losses from IBD.gaf.
  • Join our per-gene action; flag CROSS_SUBFAMILY, LOCALIZATION, NODE_LOSS, ...
  • just refresh-panther-iba-project regenerates all three tables.

Outcomes and flags

The two failure patterns

The flag is triage, not a verdict

Gene IBA term Why the cross-subfamily flag is a false positive
cdc2 CDK holoenzyme Cdc2 is the founding CDK
plo1 protein Ser/Thr kinase activity Polo kinase; conserved family-wide
rad3 DNA damage checkpoint Rad3/ATR; conserved PIKK function
ste11 dbTF activity HMG transcription factor
cdc18 replication origin binding Cdc18/Cdc6

Fine-grained PANTHER subfamilies split true orthologs across species. From REVIEW.md.

One term, two correct answers

  • GO:0000165 MAPK cascade, IBA on two kinases:
    • wis1 → ACCEPT: the MAP2K of the Sty1 stress cascade.
    • cdc7 → MARK_AS_OVER_ANNOTATED: SIN initiating kinase; the SIN is a GTPase-regulated relay, not a MAPK cascade.
  • ral2: PAINT's own IRD losses (peroxidase, cytosol, redox) on node PTN005166285 keep peroxiredoxin functions off this Kelch subfamily; the one surviving IBA is ACCEPT.

Status and next steps

  • Done: 160 IBAs analysed; written review in REVIEW.md.
  • PAINT loss table: 2,129 findings across 549 cached families; 63 IKR losses fall on a reviewed member, ready for residue reconstruction with prepare_loss_analysis.py.

Read more: projects/PANTHER_IBA_REVIEW/README.md · REVIEW.md · iba_propagation.tsv · projects/IBA_REVIEW.md