NDUFV1 curation notes

2026-09-04 — PAINT no-IBA project finishing pass (AI-assisted)

Reviewed every entry in NDUFV1-ai-review.yaml, checked the top-level description and
core_functions, and wrote the accompanying family review at
interpro/panther/PTHR11780/PTHR11780-review.yaml. The file validated clean on arrival
and still does. Status moved IN_PROGRESS → COMPLETE.

Changes made

  1. Factual error in core_functions.description: "covalently bound FMN" → non-covalently
    bound.     Complex I's FMN is a dissociable cofactor, not covalently attached — this is
    the basis of the standard FMN-release assays and of the flavin-site ROS chemistry.
    UniProt records it as a COFACTOR with the note "Binds 1 FMN" and
    ECO:0000269|PubMed:28844695, with no covalent-attachment feature. Corrected, and the
    fact that both cofactors are resolved in the human cryo-EM structure added.

  2. Two GO:0005515 IPI lines promoted KEEP_AS_NON_CORE → MARK_AS_OVER_ANNOTATED.
    The draft demoted all seven protein-binding lines uniformly, but they are not of equal
    standing. The NDUFV3 lines (PMID:24344204, PMID:30021884, PMID:33961781) and the
    RAB5IF line (PMID:31536960) reflect genuine FP-subcomplex neighbourhood and respirasome
    assembly biology, so non-core is right for them. The other two are not:

  3. PMID:32296183 (HuRI, NDUFV1–CYSRT1). A yeast two-hybrid binary screen; CYSRT1 is
    a keratinocyte-associated cytoplasmic protein with no mitochondrial role, and the
    assay places both partners outside their native compartments. The paper itself notes
    "the majority of PPIs in HuRI were found in only one screen".

  4. PMID:32814053 (Haenig et al.). ~25 unrelated partners for NDUFV1 from a single
    systematic Y2H screen designed around protein aggregation in neurodegeneration, with
    no orthogonal validation. These lines over-state what is known about NDUFV1 binding
    rather than adding non-core biology.

  5. Second core_functions entry added for GO:0009055 (electron transfer activity).
    The draft asserted electron transfer activity in the prose of the first entry's
    description but left it out of the structured molecular_function slot, so the claim
    was unindexed. Added as its own entry, matching the NEW annotation already proposed
    in existing_annotations.

  6. Empty findings: [] filled for 13 references — five GO_REF entries, seven
    Reactome reactions, and PMID:32296183 — each stating what that source actually
    contributes to this gene and, where a cached publication exists, with a verbatim
    quote. This also records why several of them are weak (inter-ontology inference
    yielding a membrane-arm process term; LONP1 reactions supporting only localization).

Bioinformatics done for this pass (reproducible)

Global BLOSUM62 alignment (BioPython PairwiseAligner, gap open −11 / extend −1) of
P49821 against P25708 (bovine), P31979 (E. coli NuoF), Q9FNN5 (Arabidopsis), O94500
(S. pombe) and Q54I90 (Dictyostelium), all fetched from the UniProt REST API; every
mapped position re-confirmed by direct indexing.

The four N3 cysteine ligands are invariant across the family:

protein N3 ligands
P49821 human NDUFV1 C379, C382, C385, C425
P25708 bovine NDUFV1 C379, C382, C385, C425
P31979 E. coli NuoF C351, C354, C357, C398
Q9FNN5 A. thaliana C402, C405, C408, C448
O94500 S. pombe C377, C380, C383, C424
Q54I90 D. discoideum C396, C399, C402, C442

Unlike the by-similarity cysteines on the partner 49 kDa subunit family (see
genes/human/NDUFS2/NDUFS2-notes.md), all four of these are structurally observed —
UniProt cites PubMed:28844695 plus PDB 5XTB/5XTD/5XTH/5XTI for each. Recorded as the
fe4s_n3_ligands site in the family review. Left at strength: CONTRIBUTES rather than
REQUIRED despite the mechanism warranting REQUIRED, because no family member has been
found that lost a ligand and the residue validator rightly refuses a REQUIRED claim with
no negative control.

The NADH-binding glycine loop (UniProt BINDING 87..96) is GRGGAGFPTG in human,
bovine, Arabidopsis and Dictyostelium, but GRGRYG at 80–85 in S. pombe O94500.
Deliberately not declared as a family motif site for that reason — a G.GG.G pattern
would over-claim. Described in prose in the family review instead.

On the "no IBA" premise — this gene gives the sharp answer

NDUFV1 does receive IBAs, but only two, and both are non-molecular-function:
GO:0045271 and GO:0006120, matching the two IBD rows at the eukaryotic node
PTN000207233. It receives no molecular-function IBA of any kind, because that node
carries no F-aspect IBD. The family's only F-aspect IBD — GO:0003954, NADH dehydrogenase
activity, dated 20260528 — sits on the bacterial node PTN000207323, seeded by E. coli
NuoF (P31979), and human NDUFV1 is not a descendant of it.

The consequence is that the defining functions of the 51 kDa flavoprotein — NADH
dehydrogenase activity, FMN binding, NAD binding, electron transfer activity — reach the
human protein only as InterPro2GO and keyword-derived IEAs, with no phylogenetic
assertion behind any of them, even though they are conserved from E. coli to human and
are what the family is named for [PMID:8288251, "plays an important role in the formation
of the NADH-binding site and is believed to be the principal site of entry for electrons
donated by NADH into the respiratory chain"].

Recommendation recorded in the family review: raising GO:0003954 (and, on the strength
of the invariant sites, GO:0010181, GO:0051287 and GO:0009055) to a node above the
bacteria/eukaryote split would be sound. Raising GO:0008137 would not — the
actinobacterial members of the same single subfamily (M. tuberculosis P9WIV6/P9WIV7,
M. bovis P65568, S. coelicolor Q9XAQ9) come from organisms whose respiratory chain
uses menaquinone, not ubiquinone. The current PAINT curation already avoids that trap by
choosing the acceptor-neutral GO:0003954; that choice should be preserved.

Why this family is coherent and its Q-module partner is not

PTHR11780 has no plastid branch: the chloroplast NDH complex has no counterpart of this
subunit at all, which is exactly why it is reduced by ferredoxin rather than by NADH.
The fetched entries slice bears this out — 0 chloroplastic members in PTHR11780 against
134 of 638 in PTHR11993, all of the latter binned into the same subfamily as human NDUFS2.
The two family reviews were written as a pair for this reason.

2026-09-17 — PR #2956 review response