Working notes for the GO annotation review in E2F1-ai-review.yaml. Sources: the
seeded GOA rows, E2F1-uniprot.txt, E2F1-deep-research-falcon.md, and the cached
publications in publications/. Quotes below are verbatim from the cited source.
Unambiguously human E2F transcription factor 1, the founding member of the activator
E2F class. The historical aliases RBAP-1 / RBBP3 record its discovery as a
retinoblastoma-protein-binding factor, which is why so much of its annotation set is
RB1-centric. UniProt features fix the domain layout: DNA_BIND 110..194, a
REGION 67..108 "Cyclin A:CDK2 binding", REGION 195..284 "Dimerization",
REGION 368..437 "Transactivation", and REGION 409..426 "RB1 binding".
The protein is an obligate heterodimer, which matters for grading nearly every MF row
[UniProt Q01094, "SUBUNIT: Component of the DRTF1/E2F transcription factor complex. Forms"].
TFDP3 is the exception that proves the rule — it binds E2F1 but the resulting dimer
cannot engage DNA PMID:17062573.
A sequence-specific Pol II transcriptional activator
[UniProt Q01094, "Transcription activator that binds DNA cooperatively with DP"], driving
the G1/S program [UniProt Q01094, "The DRTF1/E2F complex functions in the control of"].
E2F1 over-expression forces quiescent cells into S phase
PMID:20224733 and reduces the G1 fraction of cycling cells
PMID:20224733.
The deep research reaches the same conclusion
[file:human/E2F1/E2F1-deep-research-falcon.md "The best-established physiological role is
activation of a coordinated proliferation program at the G1/S transition."].
It is not an enzyme
[file:human/E2F1/E2F1-deep-research-falcon.md "Its primary molecular function is to act in
the nucleus as a sequence-specific, DP-dependent transcriptional activator."].
The Rb-E2F complex is structurally settled
PMID:16360038, with a
second, pocket-domain interface on E2F1 residues 409-426
PMID:12598654.
Repression works by masking, not by an intrinsic E2F1 repressor activity
PMID:9468140.
Decision point. GO:0000122 (negative regulation of transcription by Pol II, IMP,
PMID:20224733) was initially graded KEEP_AS_NON_CORE on the reasoning that RB1 does the
repressing. On review that was inconsistent: the same file treats the Rb-E2F corepressor
platform as core_functions[1]. E2F1 does part of the work — it supplies the
sequence-specific DNA binding that positions RB1 and its deacetylase activity on the
promoter, and repression fails when that interface is broken (the Y411C/411 mutant). This
is the scaffold shape of participation described in CLAUDE.md, so the row was upgraded to
ACCEPT. The paper is explicit that the direction depends on cell-cycle position
PMID:20224733.
Note this study used a murine fibroblast line carrying human E2F1 constructs.
Stabilisation is the trigger
PMID:12717439, and the output is apoptosis
PMID:12717439. Direct pro-apoptotic targets are well attested:
ASPP1/2 PMID:15731768, p73 via MCPH1
PMID:18660752, and the p53-independent mediator
DIP PMID:15565177.
GO:0000077 (DNA damage checkpoint signaling, IMP) was graded KEEP_AS_NON_CORE: the
checkpoint signal is generated by ATM/ATR-CHK2, and E2F1 is the transcriptional output.
The authors themselves hedge PMID:12717439. GO:0008630 was ACCEPTed as the accurate representation.
Strong, direct evidence
PMID:20972224,
PMID:20972224, and E2F1 is an accessory factor, not an enzyme
[file:human/E2F1/E2F1-deep-research-falcon.md "E2F1 is consequently an accessory
organizer, not a nuclease, ligase, helicase, or repair enzyme."].
Decision point. I considered a NEW row for GO:0006281 (DNA repair), and declined.
PMID:20972224 is already cited in GOA — for the KAT2A protein binding row — so a curator
read this paper and chose not to make a repair annotation. That is precisely the
"do not add what curators deliberately declined to add" case in CLAUDE.md. The recruitment
activity that E2F1 genuinely contributes is instead captured as MF GO:0035035
(histone acetyltransferase binding), which is where the KAT2A interaction rows were
retargeted, and core_functions[3] asserts no process term. The annotation gap is raised
in suggested_questions for expert adjudication rather than filled unilaterally.
Per the repository policy, bare protein binding takes only MODIFY / REMOVE /
UNDECIDED (a validator enforces this). Every row was traced to its partner and split:
MODIFY to GO:0001222 transcription corepressorMODIFY to GO:0046982 protein heterodimerizationMODIFY to GO:0001223: TRRAPGO:0035035; deacetylases HDAC1/SIRT1 (6 rows) →GO:0042826, including KAP1-bridged HDAC1histone deacetylase binding does not overstate this.GO:0001222; CDK8 → GO:0019901;GO:0140297.REMOVE as uninformative with no better term available: four viral E7Five Ensembl-Compara rows propagated from rat/mouse were graded
MARK_AS_OVER_ANNOTATED because they encode experimental context rather than function:
response to LPS (traceable to a microglial Reg3g GO-CAM, gocams/6348a65d00001191),
cellular response to fatty acid, hypoxia, and NGF. GO:0070345 (negative regulation of fat
cell proliferation, IEA + ISS) is separately problematic: the underlying experiments
concern blocked adipocyte differentiation via C/EBPalpha
[UniProt Q01094, "Blocks adipocyte differentiation"], and E2F1 promotes rather than
restrains proliferation — so the term looks mis-scoped at the mouse source.
GO:0045599 (negative regulation of fat cell differentiation) is properly supported
PMID:20176812 and was kept as non-core. Spermatogenesis
and glial cell proliferation were kept as non-core tissue-level outputs of the generic
G1/S program.
GO:0048255 (mRNA stabilization, IDA) was MARK_AS_OVER_ANNOTATED: E2F1 has no
RNA-binding domain, and the observation
PMID:15766563 is a downstream
phenotype most simply explained by E2F1-dependent transcription of an RNA-binding
regulator. Not removed, since the result stands unchallenged.
Four IBA rows (GO:0000978, GO:0000981, GO:0006357, GO:0035189) were all
ACCEPTed. Sequence-specific Pol II transcription regulation is the ancestral activity of
the E2F family, so the node placement is sound. GO:0035189 and GO:0006357 list
UniProtKB:Q01094 in their own WITH/FROM — per CLAUDE.md this is expected and correct, a
marker that experimental grounding exists on the target itself, and was not treated as
circular.
The largest single group of MODIFY calls (68 total) generalises or specialises to the
Pol II-specific branch: GO:0003677 DNA binding → GO:0000978; GO:0003700 /
GO:0001216 → GO:0000981 / GO:0001228; GO:0045893 / GO:0010628 / GO:0006355 →
GO:0045944 / GO:0006357; GO:0006351 (DNA-templated transcription, IEA + ISS) →
GO:0006357, since a TF regulates transcription rather than performing it;
GO:0032991 (root complex term) → GO:0090575.