RAD27-uniprot.txt, RAD27-goa.tsv: fetched via just fetch-gene yeast RAD27; 41 GOA rows seeded.publications/. Abstract-only entries includejust deep-research-falcon yeast RAD27 --fallback perplexity-liteRAD27-deep-research-asta.md, butUV_PYTHON=3.12 must be exported before any just deep-research-* recipe: uvx otherwise resolvesdeep-research-client requires >=3.12.Rad27 is the budding-yeast FEN1 orthologue: a Mg2+-dependent, structure-specific XPG/RAD2-family nuclease.
Substrate specificity is the most precisely characterized aspect. Rad27p prefers a double-flap
substrate and the 3' tail of the upstream primer, not the 5' flap, sets the cut site:
PMID:11825897
PMID:11825897
It discriminates among branch-migration isomers of the same substrate:
PMID:11825897
This is why every Rad27 product is directly ligatable and why Rad27 and Cdc9 must act in an ordered way.
Exonuclease activity. Purified as the pol alpha-associated 5'→3' exonuclease before it was recognized
as the RAD27 product:
PMID:9166764
Two primer-removal routes. Reconstitution shows the short-flap/Rad27-only route dominates, with a
minor Dna2-dependent long-flap route:
PMID:16837458
PMID:16837458
Genetically, DNA2 and RAD27 are synthetically lethal and their products co-purify:
PMID:9121462
PCNA ordering. PIP-box alleles establish that PCNA sequences Rad27 and Cdc9 entry into fragment joining:
PMID:16079237
DNA:RNA hybrids and R-loops. A 2025 study directly tested purified yeast Rad27 on synthetic R-loop
and 5' flap substrates. Wild-type Rad27 detectably cleaved R-loop 3' boundaries, but several orders of
magnitude less efficiently than flaps; Rad27-E176A failed to cleave R-loop substrates while retaining
flap cleavage. Importantly, Rad27-E176A fully suppressed DNA:RNA hybrid accumulation in Rad27-depleted
cycling cells, and EXO1 overexpression suppressed both the growth defect and hybrid accumulation,
supporting a model in which hybrids arise mainly after flap accumulation rather than because Rad27 is
the limiting direct R-loop resolvase.
PMID:40064914
PMID:40064914
Repeat instability — the strongest in vivo phenotypes.
CAG tracts: PMID:16079237
rDNA (2025, full text available): PMID:41140146 — and importantly the mechanism is not break-mediated:
PMID:41140146 with partial redundancy:
PMID:41140146
Mitochondrial pool.
PMID:19699691
GO:0005737 cytoplasm, IBA is_active_in — MODIFY to GO:0005739 mitochondrion. This is the one IBA
localization call I changed. "Active in cytoplasm" implies a cytosolic site of action, but Rad27's
substrate is DNA; the only extranuclear compartment where it has a demonstrated substrate is the
mitochondrion (which is part_of cytoplasm, so the parent is not false, just uninformative and
misleading about site of action). PTN000871783 is the current PTHR11081 fungal cytoplasm IBD and RAD27
does descend from it, but the target-specific extranuclear site is mitochondrial rather than soluble
cytosolic. Recorded with propagation_review (root_cause TERM_SCOPING_PROBLEM; failure modes
COMPARTMENT_OR_COMPLEX_MISMATCH, GRANULARITY_MISMATCH).
IBA re-review. RAD27's current GOA IBA rows all trace to PTHR11081: PTN000118792 for nucleus and
DNA replication, PTN000118612 for 5'-flap endonuclease activity, PTN000118791 for 5'-3' exonuclease
activity, PTN000871783 for cytoplasm, and PTN008960732 for DNA repair and DNA recombination. The source
for a PAINT row is the PTN ancestral node, not every extant WITH/FROM member, so the revised
propagation_review blocks use the PTN as the source entity while retaining the full GOA
supporting_entities list for traceability. DNA replication remains a TERM_SCOPING_PROBLEM to Okazaki
fragment processing; DNA recombination is defensible but non-core; the other IBA rows are core except
for cytoplasm.
GO:0005829 cytosol (IDA, PMID:22932476) — KEEP_AS_NON_CORE, not REMOVE. The source is a hypoxia
relocalization screen whose stated focus is SWI/SNF
PMID:22932476.
Rad27 is one of the surveyed nuclear proteins. I did not overrule the SGD curator — the observation is
plausible as a condition-dependent redistribution — but it is peripheral, since there is no cytosolic
substrate. The same paper's nucleus IDA was accepted outright.
GO:0007534 gene conversion at mating-type locus (IMP, PMID:10025407) — KEEP_AS_NON_CORE. Real and
specific assay, but mechanistically it reports the same lagging-strand requirement as the core Okazaki
role; the authors interpret strand invasion as creating a modified replication fork
PMID:10025407. Not a mating-type-specific function of Rad27.
GO:0006303 NHEJ (IDA, PMID:15342630) — KEEP_AS_NON_CORE. Well supported biochemically
PMID:15342630, but Rad27's contribution is 5'-end trimming, not a core identity. Parallels the
MMEJ flap-trimming role Reactome assigns to human FEN1.
PMID:9166764 title names DNA polymerase alpha, not RAD27 — annotation NOT second-guessed. The
abstract explicitly identifies the purified protein as the RTH1/RAD27 gene product by peptide sequencing,
so the 5'-3' exonuclease IDA is correctly attributed. Flagged as VERIFIED in reference_review.
GO:0005515 protein binding IPIs — REMOVE. The PCNA and Dna2 interactions are real and biologically
important, but the four IntAct rows all use the generic protein binding molecular-function term. There
is no specific PCNA-binding or Dna2-binding MF replacement term supported by the cited experiments, and
the functional consequences are already captured by the Okazaki fragment processing, primer removal and
flap endonuclease annotations. These were migrated from the legacy MARK_AS_OVER_ANNOTATED action to
REMOVE during the 2026 re-review.
One, from PMID:36672839 (full text cached), with R-loop caveats from PMID:40064914:
Hence:
The human FEN1 GO:0062176 annotation is correctly IDA — those are the human experiments.
PMID:40064914 then assayed purified S. cerevisiae Rad27 on synthetic R-loop substrates and found
detectable but very weak R-loop 3' boundary cleavage; because the same paper argues that DNA:RNA
hybrid accumulation in cycling Rad27-depleted cells primarily follows flap accumulation rather than
loss of direct R-loop cleavage, direct yeast R-loop processing should stay a question rather than a
proposed process annotation.
Several nucleus annotations originally carried quotes that did not support the compartment claimed
(most seriously, the IBA GO:0005634 is_active_in nucleus was supported by the mitochondrial
localization sentence from PMID:19699691). Rad27's nuclear localization is genuinely well established,
but it rests on dataset-level evidence — the genome-wide GFP-fusion collection (PMID:14562095) and
the curated UniProt subcellular location — not on any Rad27-specific sentence in a cached abstract.
Those entries now carry no supporting_text and say so explicitly in reason, which is more honest
than attaching a quote that is verbatim but irrelevant. PMID:19699691 is abstract-only, so its nucleus
IDA is accepted in deference to the SGD curator's full-text reading.
42 annotations (41 from GOA + 1 proposed NEW):
ACCEPT 29 · MODIFY 4 · KEEP_AS_NON_CORE 4 · REMOVE 4 · NEW 1.