Reproducible LNP1 interaction and localization provenance audit

Bottom line

Human LNP1 is UniProtKB A1A4G5 (178 aa; HGNC symbol LNP1; synonym
NP3). It is not lunapark/LNPK, UniProtKB Q9C0E8 (428 aa; HGNC symbol
LNPK; historical synonym LNP). The accessions are independently resolved in
results/uniprot_identities.tsv.

Two informative molecular-function annotations are defensible as non-core,
high-throughput/interaction-derived claims:

1. GO:0071889 14-3-3 protein binding, IPI, PMID:33961781. BioPlex recovered
   six 14-3-3 genes with LNP1 in both 293T and HCT116 cells. The raw
   accession-level protein-group rows are P27348 (YWHAQ), P31946-2 (YWHAB), P61981 (YWHAG),
   P62258 and P62258-2 (YWHAE), P63104 and the unreviewed truncated
   proteoform E7ESK7 (YWHAZ), and Q04917 (YWHAH). These identifiers record
   the source assignments; they do not establish isoform-specific detection or
   isoform-exclusive biology.
2. GO:1990782 protein tyrosine kinase binding, IPI, PMID:23503679, with
   UniProtKB:P07948 (LYN). PLATO discovery was followed by a targeted
   GST-LYN pull-down/Western validation in which Figure 2c labels LNP1.

Both annotations must retain an explicit limitation: the assays establish
physical co-association under the tested conditions, not an endogenous pathway,
substrate relationship, or purified-protein binding mechanism. AP-MS in
particular does not distinguish direct contact from a bridged/co-complex
association.

No informative GO annotation follows from the HuRI LNP1-GPRIN2 pair or the
LNP1-GAPDH crosslink. A generic protein binding annotation would add no useful
function. Complex Portal CPX-12224 is explicitly machine-learning predicted and
must not be asserted as an experimentally isolated, stable 28-member complex.

Evidence audit

BioPlex 3.0 / PMID:33961781

The official December 2019 BioPlex no-filter bait-prey tables, not only a
canonicalized interaction export, were checked. Exact source rows are retained
in
results/bioplex_LNP1_14-3-3_source_rows.tsv.

The source therefore contains six 14-3-3 genes, but eight distinct accession-level
protein-group strings across the two cell lines. E7ESK7 is a current unreviewed, 137-aa
YWHAZ sequence; it is not a seventh 14-3-3 gene. IntAct collapses the YWHAB,
YWHAE, and YWHAZ source proteoforms to canonical gene-level accessions.

The twelve IntAct records (six pairs in each cell line) use:

• MI:0007 anti tag coimmunoprecipitation;
• MI:0914 association, not MI:0915 physical association;
• MI:1060 spoke expansion;
• LNP1 as bait (MI:0496) and each 14-3-3 protein as prey (MI:0498);
• IntAct miscore 0.35.

This repeated family-selective AP-MS pattern supports an IPI annotation to
14-3-3 protein binding under normal GO interaction-evidence practice, but it
does not by itself prove direct binary binding. hu.MAP3 and AlphaFold3 add a
testable structural hypothesis, not independent direct experimental evidence.

PLATO LYN interaction / PMID:23503679

BioGRID records both stages with exact identities LYN/P07948 and LNP1/A1A4G5:

• interaction 868448: high-throughput Affinity Capture-Luminescence discovery;
• interaction 868526: low-throughput Affinity Capture-Western validation.

The full paper's Figure 2c visibly labels LNP1, shows an LNP1 band in input and
GST-LYN pull-down, and no corresponding band in the GST-Pep control. The assay
used tagged LNP1 expressed in 293T lysate and bacterially produced GST-LYN; it
supports physical association but not endogenous interaction, LYN-dependent
phosphorylation, or a physiological LYN pathway.

hu.MAP3 / PMID:40425816

The official hu.MAP3 page for huMAP3_06971.1 labels the cluster Very High
confidence and reports these LNP1 edges:

hu.MAP3 is a machine-learning integration of AP-MS, co-fractionation MS,
proximity-labeling, and related features; BioPlex is among its inputs. Its score
is an estimated co-complex/interacting probability. The paper's LNP1-YWHAE
AlphaFold3 model is predictive (reported ipTM 0.73 and pTM 0.64), not an
experimental binding validation.

There is also a residue-numbering warning. Current A1A4G5 residues 110-115 are
KFSESF. The paper states Ser114 but writes KFpSESF, which places the
phosphate on the first serine (current S112); correct S114 notation would be
KFSEpSF. The cited phosphoproteome source reports S114. Do not turn this
internally inconsistent motif typography or the AlphaFold model into a
residue-specific functional annotation.

The exact official PMID:31819260 Supplementary Table S2 row is retained in
results/raw/pmid31819260_A1A4G5_S114_source_row.tsv.
It records A1A4G5 residue 114 as serine, localization probability 0.844585, and
two spectral counts. This substantiates site detection only; it does not assign
the kinase, demonstrate 14-3-3 binding, or resolve the motif-numbering conflict.

Complex Portal CPX-12224

The official API returns:

"predictedComplex": true
"evidenceType": {"identifier": "ECO:0008004", "description": "machine-learning predicted complex"}

It cross-references huMAP3_06971.1 as an identical object, lists 28
participants (including A1A4G5 and all seven human 14-3-3 genes), and gives null
stoichiometry for every participant. This is a predicted clustering result, not
primary IPI evidence and not a basis for an in_complex: CPX-12224 assertion.

HuRI / PMID:32296183

The unresolved IntAct participant ccsb orf id: 6830 is not LNP1 and is not an
unknown protein. Official Supplementary Table 2 maps it as follows:

6830    ENST00000374317.1   ENSP00000363436.1   ENSG00000204175.5   GPRIN2
54500   ENST00000383693.7   ENSP00000373191.3   ENSG00000206535.7   LNP1

Current UniProt maps GPRIN2 to reviewed accession O60269. Supplementary Table
9 contains exactly one GPRIN2-LNP1 pair, detected in screen 1 and assay v1:

ENSG00000204175 ENSG00000206535 1   0   0   0   0   0   0   0   0   1   0   0

IntAct exposes three records for that one pair (MI:1356 validated two hybrid,
MI:1112 two hybrid prey pooling, and MI:0397 two hybrid array), all typed
MI:0915 physical association with miscore 0.56. This is credible binary Y2H
evidence, but no specific molecular function or biological process is known;
therefore it does not justify a useful GO annotation.

Nuclear XL-MS / PMID:30021884

Supplementary Table S2 contains one DSSO interlink in a TX100-insoluble U2OS
nuclear fraction. It assigns LNP1 leading residue Lys124 and GAPDH leading
residue Lys145 (the GAPDH peptide also maps to P04406-2 residue 103), score
95.05 and Q-value 0.02. IntAct represents it as A1A4G5-P04406, MI:0030
cross-linking study, MI:0915 physical association, miscore 0.40.

This is residue-proximity evidence in an intact-nucleus preparation, but it is a
single high-throughput crosslink and does not establish a selective GAPDH-binding
function, normal nuclear-speck residence, or a physiological pathway. The paper
itself highlights difficulty filtering nonspecific contacts and validating novel
PPIs. Do not add a generic protein-binding row.

Human Protein Atlas cellular-component boundary

The current HPA XML entry uses one antibody, HPA047926, with Approved (not
Enhanced/Supported) reliability. Exact cell-line calls are:

Thus vesicular staining is shared across the two assayed lines, whereas nuclear
speckles and cytosol are PC-3-only calls. These data support a cautious,
cell-line-qualified description of vesicular staining, not constitutive nuclear
speck or cytosol localization. The XML labels the call vesicles while attaching
GO:0043231, whose current GO label is the broader intracellular
membrane-bounded organelle; do not silently treat that HPA identifier as a
precise current vesicle term. Given the single Approved antibody, none of these
locations should be promoted to a core function without orthogonal validation.

Annotation recommendations

Validator-safe verbatim source text

The following strings are exact substrings of the indicated cached publication
files and can be copied without paraphrase into supporting_text.

PMID:23503679

GSTLYN precipitation and western blot analysis confirmed binding for five of seven novel candidates tested (Figure 2c).

PMID:33961781

In contrast, affinity-purification mass spectrometry (AP-MS) enables enrichment and detection of even low-abundance proteins, though exogenous expression of tagged baits is required, and extensive sample preparation has limited scalability while precluding recovery of transient interactions ( Gingras et al., 2007 ).

PMID:40425816

We also identify LNP1, an uncharacterized protein, as associated with members of the 14-3-3 complex (huMAP3_06971.1).

LNP1 has a known phosphoserine site at Ser114 (Ochoa et al, 2020) in a motif reminiscent of 14-3-3 binding (KFpSESF vs RXY/FXpSXP (Yaffe et al, 1997)).

Provided this, we used AlphaFold3 to model the interaction between LNP1 and YWHAE, a 14-3-3 subunit which had the highest hu.MAP3.0 score 0.986 to LNP1.

This provides further evidence of LNP1’s association with the 14-3-3 complex.

PMID:32296183

To map the reference interactome, we performed nine screens of Space III, followed by pairwise verification by quadruplicate retesting and sequence confirmation.

The dataset, versioned HI-III-20 (Human Interactome obtained from screening Space III, published in 2020), contains 52,569 verified PPIs involving 8,275 proteins (Supplementary Table 9).

PMID:30021884

Here we use crosslinking mass spectrometry (XL-MS) to chart the protein-protein interactions in intact human nuclei.

The major challenge is represented by the ability to identify and filter out nonspecific interactions and by challenges in validating novel observed PPIs.

HPA XML exact text

Mainly localized to vesicles. In addition localized to the nuclear speckles and cytosol.

Immunofluorescent staining of human cell line PC-3 shows localization to nuclear speckles, vesicles and cytosol.

Exact local-reference title and reusable snippets

Use this local reference identifier and title:

• id: file:human/LNP1/LNP1-bioinformatics/RESULTS.md
• title: Reproducible LNP1 interaction and localization provenance audit

The following sentences occur verbatim in this file and are suitable as
file-backed supporting text:

Official BioPlex source rows identify P31946-2, not canonical P31946, as the YWHAB protein-group accession in both 293T and HCT116 LNP1 experiments.

Official BioPlex source output contains both P62258 and P62258-2 YWHAE protein-group rows in 293T cells, while HCT116 contains P62258.

IntAct classifies the BioPlex LNP1-14-3-3 records as MI:0007 anti tag coimmunoprecipitation, MI:0914 association, and MI:1060 spoke expansion rather than direct binary binding.

BioGRID maps the validated PLATO pair to LYN/P07948 and LNP1/A1A4G5 and classifies interaction 868526 as Affinity Capture-Western.

Official HuRI supplements map CCSB ORF 6830 to GPRIN2/O60269 and ORF 54500 to LNP1/A1A4G5, yielding one GPRIN2-LNP1 binary Y2H pair rather than self-binding.

Complex Portal marks CPX-12224 as predictedComplex true with ECO:0008004 machine-learning predicted complex evidence and unknown participant stoichiometry.

Human Protein Atlas antibody HPA047926 has Approved reliability and reports vesicles in PC-3 and U2OS, with nuclear-speck and cytosolic signal only in PC-3.

Official PMID:31819260 Supplementary Table S2 records A1A4G5 Ser114 with localization probability 0.844585 and two spectral counts.

Reproducibility and provenance

Run:

cd genes/human/LNP1/LNP1-bioinformatics
just check

The workflow uses Python's standard library, reads identifiers/URLs from
config.json, writes direct and filtered results under results/, records URL,
HTTP metadata, byte ranges, size, and SHA-256 for every retrieved source in
results/source_manifest.tsv, and fails when an
expected identity or edge changes.

• [x] Source inputs, URLs, target identifiers, and the output directory come
  from CLI arguments or config.json; report filenames are deliberately
  fixed for this gene-specific LNP1 audit.
• [x] Tested an explicit confounder/control input: Q9C0E8 resolves to LNPK, not
  LNP1.
• [x] All requested interaction, complex, and localization analyses completed.
• [x] Direct API/database results and exact supplement rows are retained under
  results/ and results/raw/.
• [x] Detailed source provenance and checksums are retained in the manifest.