ISOFORMS

Genes whose splice isoforms or cleavage products do different, sometimes opposite, things

AI Gene Review · projects/ISOFORMS · 2026

Bottom line

  • Most GO annotation is gene-level, but Bcl-xL/Bcl-xS and α-MSH/β-endorphin come from one gene with opposite effects.
  • We reviewed 2,774 annotations on 16 paradigm genes and added isoform, negated and functional_isoforms to the data model.
  • Conflation is common: 246 rows marked over-annotated, e.g. 8 in BCL2L1, 21 in VEGFA, 63 in mouse App.

The problem

The gene set

Tier Genes Contrast
1 · paradigms AGRN, WT1, BCL2L1, FAS, VEGFA, CASP9 antagonistic or dominant-negative isoforms
2 · tissue / development FN1, TPM1, TPM3, DSCAM EDA/EDB, muscle vs cytoskeletal
3 · further cases FGFR2, PKM, STAT3 IIIb/IIIc ligands, PKM1/PKM2, STAT3α/β
Polyproteins POMC, APP (human), App (mouse) cleavage products

DSCAM caveat: human DSCAM has 2 isoforms, not the 38,016 of fly Dscam1.

Recording distinct products: POMC

functional_isoforms on the POMC review: each cleavage product with its UniProt chain (PRO_…, not the PRO ontology) and its own terms. Two of five products shown.

Over-annotation by gene

Findings

  1. BCL2L1: pro-apoptotic Bcl-xS function attached to the gene whose main product is anti-apoptotic.
  2. FAS: GOA carries both positive and negative regulation of apoptosis; membrane vs soluble decoy isoforms explain it.
  3. VEGFA: VEGF165B is anti-angiogenic while canonical isoforms are pro-angiogenic.
  4. App/APP: splicing (KPI domain) and cleavage (sAPPα vs Aβ) in one gene.
  5. An isoform tag records what was tested, not necessarily what is unique to that isoform.

Status and next steps

  • 13 of 16 reviews COMPLETE; WT1 and VEGFA IN_PROGRESS, DSCAM DRAFT.
  • Not started: PTBP1/2, MST1R.
  • Only 12 reviews repo-wide use functional_isoforms so far.

Read more: projects/ISOFORMS.md · projects/ISOFORMS/genes.csv · src/ai_gene_review/schema/gene_review.yaml