C5orf46 encodes an 87-residue precursor that is cleaved after a 23-residue signal peptide to release a 64-residue secreted peptide. The mature peptide is unusual for a human host-defence molecule: it is anionic rather than cationic, with sixteen acidic residues against ten basic ones and an isoelectric point of 4.54, it contains no cysteines and so forms no disulphide-stabilised fold, and it is amphipathic, pairing a highly acidic and structurally disordered aspartate- and proline-rich N-terminal half with a central hydrophobic segment predicted to be helical, the peptide's helical content having been confirmed by circular dichroism without localising it to that segment. Expression is concentrated in skin, salivary gland and blood vessel, which is the origin of the gene's alternative name skin and saliva secreted protein 1, and the protein itself has been detected in human plasma by mass spectrometry and catalogued in the exosome fraction of parotid saliva. Purified recombinant mature peptide kills Gram-negative bacteria - Escherichia coli including the O157:H7 serotype, Vibrio cholerae and Pseudomonas aeruginosa - in a concentration-dependent manner with measured minimum inhibitory concentrations, while leaving the Gram-positive Staphylococcus aureus and Listeria monocytogenes and yeast unaffected; electron microscopy of treated bacteria shows the cells becoming irregular and collapsing. Administering the peptide to mice after a lethal Escherichia coli O157:H7 challenge raises survival substantially. The mouse orthologue behaves the same way. How the peptide kills is unresolved: the growth kinetics are those of an agent acting on the cell envelope, but an intracellular target has not been excluded, no molecular target has been identified, and the peptide is also cytotoxic at micromolar concentrations to mammalian cells generally, including normal T lymphocytes, keratinocytes and fibroblasts as well as lymphoma lines. Beyond this antibacterial activity the peptide has no characterised molecular function, no established binding partner and no known receptor, and whether the endogenous peptide is required for host defence at the body surfaces where it is made has not been tested.
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0005576 extracellular region | IEA GO_REF:0000044 | ACCEPT | Summary: Accepted as a core location. The evidence route is weak - a mapping from a curator inference that itself rests on a predicted signal peptide - but the conclusion is independently measured three ways that are not in that route. Reason: Taken on its own the provenance of this row would deserve caution. Its input is UniProtKB-SubCell:SL-0243 'Secreted', which comes from the entry's SUBCELLULAR LOCATION line flagged ECO:0000305, a curator inference; and that inference rests on SIGNAL 1..23 flagged ECO:0000255, a prediction. So the chain from evidence to annotation contains no measurement at any step. What rescues it is that the conclusion has been reached separately by routes the annotation does not cite. The protein is identified in human plasma by mass spectrometry using an assay that enriches exactly the 2-10 kDa window the 7.22 kDa mature peptide occupies. It is catalogued in the exosome fraction of parotid saliva, which is the basis of this gene's other cellular-component row. UniProt records PE 1: Evidence at protein level with a Proteomics identification keyword. HGNC's alias name for the gene is skin and saliva secreted protein 1 and the Human Protein Atlas calls its expression group-enriched in blood vessel, salivary gland and skin - both consistent with a secreted product of surface tissues. And the peptide's assayed activity is exerted on bacteria outside the cell, which is only coherent for an extracellular protein. The route by which GOA asserted this term is therefore weaker than the term itself, and both facts are worth recording. One further point on the reference: GO_REF:0000044 is the UniProt SubCell mapping and is very much live, returning 139,714 human annotations, in contrast to GO_REF:0000043, the Swiss-Prot keyword reference, which now returns 0 - so this row should not be mistaken for a retired keyword-derived annotation. Supporting Evidence: PMID:31308252 Applying this method to perform deep-proteome profiling of human plasma we identify C5ORF46 as a previously uncharacterized human plasma protein. file:human/C5orf46/C5orf46-uniprot.txt CC -!- SUBCELLULAR LOCATION: Secreted {ECO:0000305}. file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 file:human/C5orf46/C5orf46-uniprot.txt PE 1: Evidence at protein level; file:human/C5orf46/C5orf46-uniprot.txt DR HPA; ENSG00000178776; Group enriched (blood vessel, salivary gland, skin). file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | isoelectric point | `PI = 4.54` | **4.54** | |
| GO:0005515 protein binding | IPI PMID:28514442 Architecture of the human interactome defines protein commun... | MARK AS OVER ANNOTATED | Summary: Over-annotated. An affinity-proteomics association with TMBIM6 (UniProtKB:P55061) from the BioPlex 2.0 dataset, with no orthogonal validation and no functional consequence measured. Reason: TMBIM6 (BI-1) is a six-pass ER membrane protein with 132 distinct IntAct partners. This is the one row in the set that is at least topologically coherent, and that is worth stating rather than flattening: C5orf46 is the tagged bait in both BioPlex records, so this is a pull-down of an over-expressed signal-peptide protein that genuinely transits the ER, recovering an abundant resident of the compartment it passes through. That makes it a plausible encounter rather than a topologically impossible one - and also exactly the kind of encounter that co-residence in the ER produces without any functional relationship. This row is the BioPlex 2.0 record of the TMBIM6 association, detected by anti tag coip in HEK293T. The same partner appears in the other BioPlex release, so the two GOA rows for TMBIM6 are one affinity-proteomics method in two releases of one programme rather than two independent assays. Both carry MI-score 0.56. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against PEX12 (UniProtKB:O00623) from one screen, carrying no functional information about this peptide. Reason: PEX12 is a five-pass peroxisomal membrane protein and a component of the peroxisomal protein import machinery. Peroxisomes import their contents from the cytosol, so a signal-peptide-directed secretory peptide has no route to this membrane. PEX12 is also the least promiscuous of the eleven Swiss-Prot membrane partners at 37 distinct IntAct partners, which is noted because it weakens the hub explanation for this particular row even though the single-screen and hydrophobicity arguments still apply. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against SGTA (UniProtKB:O43765) from one screen, carrying no functional information about this peptide. Reason: SGTA is one of the two non-membrane partners, and it is the most diagnostic member of the set. Its curated function is binding exposed hydrophobic segments: UniProt describes it as a co-chaperone that binds hydrophobic-patches-containing clients and that binds the transmembrane domain of newly synthesized proteins, acting in the ASNA1/BAG6 module that triages mislocalised membrane proteins. This is the partner that carries the curated hydrophobic-client function - SGTB, its paralogue, does not - so the hydrophobicity reading rests on this row specifically. Recovering the cell's own detector of exposed transmembrane helices, from a bait whose most hydrophobic window is an uncleaved signal peptide, is what a hydrophobicity artefact looks like. It is also a hub, with 281 distinct IntAct partners. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against TBXA2R (UniProtKB:Q0VAB0) from one screen, carrying no functional information about this peptide. Reason: The partner accession here is not the canonical protein. Q0VAB0 is an unreviewed 259-residue TrEMBL entry named for TBXA2R, whereas canonical human TBXA2R is Q0VAB0's reviewed counterpart P21731 at 343 residues - an 84-residue truncation. So even read at face value this row does not record an interaction with the thromboxane A2 receptor as curated. Its 13 distinct IntAct partners are fewer than the subject's 14, so no promiscuity argument applies to this row either; what remains is a single-screen Y2H hit against a partial clone. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against AQP6 (UniProtKB:Q13520) from one screen, carrying no functional information about this peptide. Reason: AQP6 is a six-pass aquaporin of intracellular vesicles in the renal collecting duct. It is the most promiscuous protein in the set by a wide margin, with 382 distinct IntAct partners against the subject's 14. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against EBP (UniProtKB:Q15125) from one screen, carrying no functional information about this peptide. Reason: EBP is a four-pass ER membrane sterol isomerase of cholesterol biosynthesis, with 240 distinct IntAct partners. Nothing in this gene's literature connects it to sterol metabolism. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against LHFPL5 (UniProtKB:Q8TAF8) from one screen, carrying no functional information about this peptide. Reason: LHFPL5 is a four-pass tetraspan membrane protein of cochlear hair-cell mechanotransduction, with 138 distinct IntAct partners. Nothing connects it to a salivary and cutaneous secreted peptide. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against SGTB (UniProtKB:Q96EQ0) from one screen, carrying no functional information about this peptide. Reason: SGTB is the second non-membrane partner and SGTA's paralogue, a co-chaperone that binds HSC70/HSP70 and regulates their ATPase activity, with 202 distinct IntAct partners. Its own UniProt FUNCTION makes no hydrophobic-client claim - that is curated for SGTA, not for SGTB - so this row is not independent evidence for the hydrophobicity reading, only consistent with it. What it does add is that the screen recovered both human members of the SGT family, a pair 55.6% identical with the same TPR architecture, which is hard to explain by any pathway a secreted antibacterial peptide participates in. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against RUSF1 (UniProtKB:Q96GQ5) from one screen, carrying no functional information about this peptide. Reason: RUSF1 is a single-pass membrane protein of largely unknown function, with 199 distinct IntAct partners. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against TMEM80 (UniProtKB:Q96HE8) from one screen, carrying no functional information about this peptide. Reason: TMEM80 is a four-pass membrane protein annotated to cilium and membrane, with 66 distinct IntAct partners. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against MIMS2 (UniProtKB:Q96KR6) from one screen, carrying no functional information about this peptide. Reason: MIMS2 (formerly FAM210B) is a two-pass mitochondrial inner-membrane protein, with 101 distinct IntAct partners. The mitochondrial matrix side of that membrane is not reachable by a secretory-pathway peptide. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against SLC30A2 (UniProtKB:Q9BRI3) from one screen, carrying no functional information about this peptide. Reason: SLC30A2 (ZnT2) is a five-pass zinc transporter of secretory vesicles and endolysosomes, with 173 distinct IntAct partners. This is the partner whose UniProt entry reads NbExp=5, and the five records are still one publication and the same three sub-methods, two of them logged twice - so the higher number is not higher-quality evidence. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:32296183 A reference map of the human binary protein interactome. | MARK AS OVER ANNOTATED | Summary: Over-annotated. A single yeast two-hybrid hit against TIMMDC1 (UniProtKB:Q9NPL8) from one screen, carrying no functional information about this peptide. Reason: TIMMDC1 is a four-pass mitochondrial membrane assembly factor for respiratory complex I, with 185 distinct IntAct partners. This row is one of twelve that come from a single publication, the HuRI binary interactome (PMID:32296183). IntAct logs each of those twelve partners under three names of the same yeast two-hybrid screen - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - which is why the UniProt entry reads NbExp=3 for twelve partners and NbExp=5 for SLC30A2, where two of the three sub-methods are logged twice. The number counts sub-methods, not experiments. All twelve share one MI-score, 0.56, and no orthogonal assay supports any of them: there is no co-immunoprecipitation of endogenous protein, no biophysical binding measurement and no reciprocal pull-down anywhere in the record. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0005515 protein binding | IPI PMID:33961781 Dual proteome-scale networks reveal cell-specific remodeling... | MARK AS OVER ANNOTATED | Summary: Over-annotated. An affinity-proteomics association with TMBIM6 (UniProtKB:P55061) from the BioPlex 3.0 dataset, with no orthogonal validation and no functional consequence measured. Reason: TMBIM6 (BI-1) is a six-pass ER membrane protein with 132 distinct IntAct partners. This is the one row in the set that is at least topologically coherent, and that is worth stating rather than flattening: C5orf46 is the tagged bait in both BioPlex records, so this is a pull-down of an over-expressed signal-peptide protein that genuinely transits the ER, recovering an abundant resident of the compartment it passes through. That makes it a plausible encounter rather than a topologically impossible one - and also exactly the kind of encounter that co-residence in the ER produces without any functional relationship. This row is the BioPlex 3.0 record of the TMBIM6 association, detected by anti tag coip in HEK293T. The same partner appears in the other BioPlex release, so the two GOA rows for TMBIM6 are one affinity-proteomics method in two releases of one programme rather than two independent assays. Both carry MI-score 0.56. Set-level context, restated on every row so each can be read alone; the full census with its tables is in the gene's bioinformatics RESULTS.md, sections C and D. All fourteen rows come from two pipelines, and the partner set has a single shared property. Eleven of the thirteen partners are integral membrane proteins, carrying 44 annotated transmembrane segments between them; the other two are SGTA and SGTB, the two human members of the SGT family of TPR co-chaperones, and the hydrophobic-client function is curated for SGTA only - SGTB's entire UniProt FUNCTION is HSC70/HSP70 ATPase regulation, so it is consistent with the reading rather than a second curated instance, and that asymmetry is computed from the cached UniProt records rather than asserted (check H, four directions break-tested). C5orf46 is the bait in 26 of the 38 two-hybrid records, and that bait is the full open reading frame, whose most hydrophobic 19-residue window (Kyte-Doolittle +2.51) is the uncleaved signal peptide - more hydrophobic than anything in the mature chain (+1.08). A bait presenting the most hydrophobic segment in the construct, recovering a partner set whose only shared property is hydrophobicity, is better explained by the screen's design than by the peptide's biology. Twelve of the thirteen partners also exceed the subject's 14 distinct IntAct partners, up to AQP6 at 382, with PEX12 at 37 and the truncated TBXA2R clone at 13 recorded as the exceptions rather than smoothed away. And every interaction in the set was scored either in the Saccharomyces cerevisiae nucleus or in a HEK293T lysate, so no partner was tested in its native topology. Marked as over-annotated rather than removed. These are genuine records from a real screen and no positive argument is available that the interaction is false; what is absent is replication, orthogonal validation, and any functional consequence. GO:0005515 is also uninformative by construction, and no better molecular function term can be substituted, because nothing about the partner tells us what the peptide does. Supporting Evidence: file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | PMID:32296183 (HuRI) | `two hybrid array` | 13 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTA | O43765 | Swiss-Prot | 313 | 0 | Cytoplasm; Nucleus | 281 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | SGTB | Q96EQ0 | Swiss-Prot | 304 | 0 | (none stated) | 202 | 1 | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | **SGTB** `Q96EQ0` | **none** | Belongs to the SGT family | file:human/C5orf46/C5orf46-uniprot.txt FT SIGNAL 1..23 |
| GO:0070062 extracellular exosome | HDA PMID:19199708 Proteomic analysis of human parotid gland exosomes by multid... | KEEP AS NON CORE | Summary: Kept as a non-core location. The detection is real and it corroborates secretion, but it records where the peptide was found in one proteomic survey, not where it does its work. Reason: The protein was catalogued in the exosome fraction of human parotid saliva by shotgun proteomics, and the detection is coherent rather than incidental: HGNC's alias name for the gene is skin and saliva secreted protein 1, the Human Protein Atlas calls expression group-enriched in blood vessel, salivary gland and skin, and the only functional paper reports the same salivary-gland and skin distribution. Verified as a descendant of GO:0005576, so this row and the extracellular-region row are consistent rather than competing, with this one strictly more specific. It is retained for that reason. Non-core rather than core, on two grounds. The first is what the evidence is: a single HDA detection in one large-scale dataset, which the HDA code already encodes as such, and an exosome preparation from a secretory fluid co-purifies abundant soluble secreted proteins routinely. The second is decisive, and it is about what a core location asserts. Every measurement of what this peptide does was made on free peptide added to a bacterial culture or injected into an animal, and the paper's own SUMO-fusion control - activity abolished by an N-terminal tag and restored on cleavage - argues that the functional species is free soluble peptide rather than peptide packaged in a vesicle. Nothing shows the peptide acting in or on an exosome lumen. Listing it among core locations would therefore assert a compartment of action that no experiment supports, so the core location is GO:0005576 alone. The wider practice agrees: across the merged human reviews GO:0070062 is kept non-core or marked over-annotated about eight times for every acceptance, and it appears in core_functions.locations in only two genes, both with genuine exosome biology. One limitation stated plainly: the cached copy of this paper does not include the protein catalogue in which C5orf46 appears, so no sentence of it is quoted about this gene and the row is assessed on its provenance. That provenance was checked, and the check came back clean rather than damning. Fully paginated, the reference carries 396 annotations over 396 distinct entities, all to GO:0070062, all HDA, all assigned by UniProt - one entity, one annotation, one localisation term - and no functional or phenotype term appears anywhere in it. That is a proteomic inventory, not the pattern in which one experiment's phenotype is projected across a set of proteins. Supporting Evidence: file:human/C5orf46/C5orf46-uniprot.txt DR GO; GO:0070062; C:extracellular exosome; HDA:UniProtKB. file:human/C5orf46/C5orf46-uniprot.txt DR HPA; ENSG00000178776; Group enriched (blood vessel, salivary gland, skin). PMID:33804835 First, the mRNA expression of AP-64 in the skin and salivary gland was discovered using the TCGA database PMID:33804835 SUMO-AP-64 was expressed in a soluble form but failed to inhibit the growth of DH5Ξ± cells. After removal of the SUMO tag, AP-64 exhibited strong antibacterial effects. file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md **Annotations in all of GOA citing `PMID:33804835`, the sole functional characterisation of |
| GO:0050829 defense response to Gram-negative bacterium | IDA PMID:33804835 AP-64, Encoded by C5orf46, Exhibits Antimicrobial Activity a... | NEW | Summary: Proposed as a new annotation. The gene's only functional characterisation has never been curated into GO for any organism, and this term is the one its measurements support most precisely. Reason: The absence, not any existing row, is this gene's dominant annotation defect. Purified recombinant peptide kills four Gram-negative species with a concentration dependence spanning 0.1 to 10 micromolar and measured minimum inhibitory concentrations; electron microscopy shows the treated bacteria losing shape and disintegrating; and administering the peptide to mice after a lethal Escherichia coli O157:H7 challenge raises 24-hour survival from 20 per cent to 90 per cent. None of this is in GOA: a QuickGO query by reference returns zero annotations citing PMID:33804835 anywhere, against a positive control of 396 for PMID:19199708, so the endpoint works and the absence is real. The specific term is chosen over its parent GO:0042742 because the discrimination was measured in both arms rather than assumed: the same assay found no appreciable activity against the Gram-positive pathogens Staphylococcus aureus and Listeria monocytogenes, and none against yeast. Ancestry was fetched rather than inferred from the labels - GO:0042742 is in GO:0050829's is_a/part_of closure - so the specific term entails the general one and nothing is given up by using it. There is an exact curatorial precedent, and it is what turns this from a request into a consistency argument. The same group had previously characterised C10orf99 as a human antimicrobial peptide under the same naming convention, AP-57, and the AP-64 paper cites that work as its own predecessor. C10orf99 (UniProtKB:Q6UWK7, GPR15LG) was curated: UniProt gave it the alternative name 'Antimicrobial peptide with 57 amino acid residues', and GOA carries four IDA annotations from the AP-57 paper, among them GO:0050830 defense response to Gram-positive bacterium and GO:0050832 defense response to fungus. The two peptides are near-mirror images: AP-57 is basic with four cysteines and kills Gram-positives and a fungus, while AP-64 is anionic with none and kills Gram-negatives but neither Gram-positives nor yeast. So GO:0050829 for this gene is the exact counterpart of the GO:0050830 that the sibling already holds, from an equivalent experiment in an equivalent paper by the same laboratory. The class is demonstrably curatable and has been curated; this gene was simply missed. The molecule assayed is this gene's physiological product, not a synthetic fragment, and that was checked rather than assumed: recomputing from UniProt's annotated CHAIN 24..87 gives 64 residues, 7.22 kDa, isoelectric point 4.54 and zero cysteines, reproducing all four properties the paper states for AP-64, with the same mass routine reproducing UniProt's stated 9693 Da for the 87-residue precursor as a positive control. The direction of the evidence needs stating because GO's evidence codes do not encode it. Every experiment is exogenous addition of purified peptide - to bacterial cultures in vitro, and by intraperitoneal injection 30 minutes after a lethal challenge in vivo - so what is established is that the peptide is sufficient to kill Gram-negative bacteria and sufficient to protect an animal when administered. No knockout, knockdown or patient loss-of-function has been challenged with a pathogen in either species, so the gene has not been shown to be required for antibacterial defence. No molecular function term is proposed alongside these process terms, and that is deliberate. The authors leave the mechanism open between cell-envelope damage and an intracellular target and say so twice, once in interpreting the growth curve and again in their limitations. The alpha-helix is a PSIPRED prediction with circular-dichroism support for helical content, which establishes that a helix exists, not that it forms a pore or permeabilises a bilayer. Deriving a pore-forming or membrane-disrupting activity from it would be reading a structure prediction as a mechanism. GO offers no term for 'kills bacteria by an unidentified route', and inventing one would be an over-annotation in the opposite direction. Two neighbouring terms are declined for a reason that belongs in the record rather than only in the notes, because declining them while accepting GO:0050829 from the same experiments would otherwise look inconsistent. GO:0019731 antibacterial humoral response and GO:0061844 antimicrobial humoral immune response mediated by antimicrobial peptide both require the response to occur in a BODY FLUID. The peptide is present in plasma and in saliva, but every assay was done in culture medium or by injecting peptide into an animal, so no experiment shows the endogenous peptide mounting a response in either fluid; those two terms would convert a sufficiency result into an organism-level process claim, which GO:0050829 and GO:0031640 do not. GO:0140367 antibacterial innate immune response is declined on its own differentia, which requires components that directly recognise a pathogen component - no target has been identified here. Supporting Evidence: PMID:33804835 AP-64 exhibited significant antibacterial activity against Gram-negative bacteria, including Escherichia coli DH5Ξ±, Escherichia coli O157:H7, Vibrio cholerae, and Pseudomonas aeruginosa. PMID:33804835 AP-64 and Gm94 lacked appreciable bactericidal activity against the pathogenic Gram-positive bacteria Staphylococcus aureus and Listeria monocytogenes PMID:33804835 AP-64 and Gm94 effectively inhibited the growth of bacteria in the concentration range of 0.1β10 ΞΌM PMID:33804835 SEM examination revealed that a 2-h treatment with the proteins at the concentration of 10 ΞΌM caused an apparent irregularity and collapse in cell shape. PMID:33804835 The survival rate of mice treated with PBS was 20% at 24 h after bacterial inoculation, whereas the survival rates of mice injected with AP-64 and Gm94 (500 ΞΌg/kg of body weight) were 90% and 80%, respectively PMID:33804835 SUMO-AP-64 was expressed in a soluble form but failed to inhibit the growth of DH5Ξ± cells. After removal of the SUMO tag, AP-64 exhibited strong antibacterial effects. file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | length | 64 residues (the "64" in the name) | **64** | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | molecular weight | `MW = 7.2` | **7.22 kDa** | file:human/C5orf46/C5orf46-deep-research-affinage.md Beyond its antibacterial role, no defined molecular mechanism, receptor, or binding partner has been characterized in the available corpus. PMID:33804835 we also observed that the yeast cells were unsusceptible to the peptide treatment |
| GO:0031640 killing of cells of another organism | IDA PMID:33804835 AP-64, Encoded by C5orf46, Exhibits Antimicrobial Activity a... | NEW | Summary: Proposed as a new annotation, and not redundant with the defense-response row: both closures were fetched and neither term is an ancestor of the other. Reason: This is the second, non-overlapping description of the same experiment, and the check that establishes it is not redundant was done rather than assumed. GO:0031640 is not a descendant of GO:0042742 - it sits under GO:0001906 cell killing, in a different branch from the defence-response hierarchy - so a gene annotated to GO:0050829 does not thereby hold GO:0031640, and the two rows say different things: one that the peptide participates in a response protecting the organism against Gram-negative bacteria, the other that it kills the cells of another organism. The direct evidence is the same: colony-forming-unit loss and measured minimum inhibitory concentrations against four Gram-negative species, with electron microscopy showing the bacteria collapsing. Curatorial precedent supports the pairing - dermcidin, the anionic cysteine-free human antimicrobial peptide the paper itself names as its closest comparator, carries GO:0031640 alongside its defence-response annotations - though the query behind that, now committed rather than asserted, shows dermcidin holds GO:0031640 by InterPro2GO IEA (GO_REF:0000002) rather than by a curator's experimental call, so it is precedent that the term is used for this class and not that it was assigned from an experiment. Dermcidin's own experimental defence rows are GO:0042742 and GO:0140367, both IDA. The non-redundancy argument therefore rests on the verified closure fetch above and not on this precedent, which is offered as corroboration at its true strength. The molecule assayed is this gene's physiological product, not a synthetic fragment, and that was checked rather than assumed: recomputing from UniProt's annotated CHAIN 24..87 gives 64 residues, 7.22 kDa, isoelectric point 4.54 and zero cysteines, reproducing all four properties the paper states for AP-64, with the same mass routine reproducing UniProt's stated 9693 Da for the 87-residue precursor as a positive control. The direction of the evidence needs stating because GO's evidence codes do not encode it. Every experiment is exogenous addition of purified peptide - to bacterial cultures in vitro, and by intraperitoneal injection 30 minutes after a lethal challenge in vivo - so what is established is that the peptide is sufficient to kill Gram-negative bacteria and sufficient to protect an animal when administered. No knockout, knockdown or patient loss-of-function has been challenged with a pathogen in either species, so the gene has not been shown to be required for antibacterial defence. No molecular function term is proposed alongside these process terms, and that is deliberate. The authors leave the mechanism open between cell-envelope damage and an intracellular target and say so twice, once in interpreting the growth curve and again in their limitations. The alpha-helix is a PSIPRED prediction with circular-dichroism support for helical content, which establishes that a helix exists, not that it forms a pore or permeabilises a bilayer. Deriving a pore-forming or membrane-disrupting activity from it would be reading a structure prediction as a mechanism. GO offers no term for 'kills bacteria by an unidentified route', and inventing one would be an over-annotation in the opposite direction. Two neighbouring terms are declined for a reason that belongs in the record rather than only in the notes, because declining them while accepting GO:0050829 from the same experiments would otherwise look inconsistent. GO:0019731 antibacterial humoral response and GO:0061844 antimicrobial humoral immune response mediated by antimicrobial peptide both require the response to occur in a BODY FLUID. The peptide is present in plasma and in saliva, but every assay was done in culture medium or by injecting peptide into an animal, so no experiment shows the endogenous peptide mounting a response in either fluid; those two terms would convert a sufficiency result into an organism-level process claim, which GO:0050829 and GO:0031640 do not. GO:0140367 antibacterial innate immune response is declined on its own differentia, which requires components that directly recognise a pathogen component - no target has been identified here. Supporting Evidence: PMID:33804835 AP-64 exhibited significant antibacterial activity against Gram-negative bacteria, including Escherichia coli DH5Ξ±, Escherichia coli O157:H7, Vibrio cholerae, and Pseudomonas aeruginosa. PMID:33804835 AP-64 and Gm94 lacked appreciable bactericidal activity against the pathogenic Gram-positive bacteria Staphylococcus aureus and Listeria monocytogenes PMID:33804835 AP-64 and Gm94 effectively inhibited the growth of bacteria in the concentration range of 0.1β10 ΞΌM PMID:33804835 SEM examination revealed that a 2-h treatment with the proteins at the concentration of 10 ΞΌM caused an apparent irregularity and collapse in cell shape. PMID:33804835 The survival rate of mice treated with PBS was 20% at 24 h after bacterial inoculation, whereas the survival rates of mice injected with AP-64 and Gm94 (500 ΞΌg/kg of body weight) were 90% and 80%, respectively PMID:33804835 SUMO-AP-64 was expressed in a soluble form but failed to inhibit the growth of DH5Ξ± cells. After removal of the SUMO tag, AP-64 exhibited strong antibacterial effects. file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | length | 64 residues (the "64" in the name) | **64** | file:human/C5orf46/C5orf46-bioinformatics/RESULTS.md | molecular weight | `MW = 7.2` | **7.22 kDa** | file:human/C5orf46/C5orf46-deep-research-affinage.md Beyond its antibacterial role, no defined molecular mechanism, receptor, or binding partner has been characterized in the available corpus. |
Loading supporting contentβ¦
Download this section (compressed HTML)Q: Should GO and MGI curate PMID:33804835? It is the sole functional characterisation of C5orf46 in any organism, and a QuickGO query by reference returns zero annotations citing it anywhere in GOA - against 396 for the control reference PMID:19199708, so this is a real absence and not a failed query. The consequence is visible on the mouse gene: MGI records Gm94 (Q3V2D2) with the ND 'no biological data available' root term in all three aspects, while that paper reports Gm94 itself killing Gram-negative bacteria in vitro and protecting mice in vivo. Curating the mouse rows would additionally open an ISO/ISS route to the human gene which cannot exist today, because transfer needs a source annotation and there is none anywhere in the family. What makes this a consistency question rather than a wish is that the same laboratory's previous paper on the same kind of molecule WAS curated: C10orf99, which they named AP-57 and which the AP-64 paper cites as its own predecessor, carries GO:0050830 and GO:0050832 by IDA from that paper (PMID:25585381), and UniProt even took 'Antimicrobial peptide with 57 amino acid residues' into the entry as an alternative name. AP-64 is the anionic, Gram-negative-killing counterpart of that basic, Gram-positive-killing peptide, so GO:0050829 is the mirror of a term the sibling already holds. The class is curatable, has been curated, and this gene was missed.
Suggested experts: GO Central curators, MGI curators, UniProt curators
Q: Should the UniProt entry for Q6UWT4 still read 'Uncharacterized protein C5orf46' with no FUNCTION comment? The entry's own reference list stops at four sequencing papers, and PMID:33804835 - which purified the mature peptide, measured minimum inhibitory concentrations against four Gram-negative species, imaged the killed bacteria, and protected mice with it - is absent. The mature-chain properties the paper states are exactly those of the entry's annotated CHAIN 24..87 (64 residues, 7.22 kDa, pI 4.54, no cysteines), so the identification is not in doubt. A FUNCTION comment, an Antimicrobial or Antibiotic keyword, and the missing reference would all be supportable. The same group's earlier peptide is the precedent here too: UniProt took 'Antimicrobial peptide with 57 amino acid residues' into Q6UWK7 as an alternative name on the strength of an equivalent paper, so the treatment requested is one this database has already applied to the sibling molecule. Raised as a UniProt correction rather than as a GO action, since there is no GO row to act on.
Suggested experts: UniProt curators
Q: Does GO need a term for antimicrobial activity whose mechanism is not established? For this gene the killing is well measured and the route is explicitly unresolved between cell-envelope damage and an intracellular target. The available molecular-function options all assert a mechanism - pore formation, membrane disruption, a specific binding activity - so the correct action here is to propose no molecular function at all, which leaves the gene's best-established property unrepresented in the molecular-function aspect. Is that the intended behaviour, or is there a gap? The question generalises to the many antimicrobial peptides whose mechanism is inferred from growth kinetics rather than demonstrated.
Suggested experts: GO Central ontology editors
Q: Should a bare GO:0005515 row be created for every partner in a single high-throughput screen, when the screen's own metadata says the records are not independent? Twelve of this gene's fourteen protein-binding rows come from one HuRI screen, logged in IntAct under three names of that screen, which is why UniProt reads NbExp=3 for twelve partners and NbExp=5 for one. The partner set has a single shared property - eleven of thirteen are integral membrane proteins and the other two are SGTA and SGTB, the co-chaperones that bind exposed transmembrane helices - while the bait is a full open reading frame whose most hydrophobic window is an uncleaved signal peptide. One partner accession, Q0VAB0, is not even the canonical protein but an unreviewed 259-residue clone of a 343-residue receptor. Would a promiscuity or topology-compatibility filter at import be worth considering?
Suggested experts: IntAct curators, GOA curators
Q: Is there a mechanistic account of an anionic antimicrobial peptide acting on Gram-negative bacteria? The cationic-peptide model that explains the defensins and LL-37 cannot apply to a peptide with sixteen acidic and ten basic residues and a pI of 4.54. The paper notes dermcidin as a precedent for the charge and the absence of cysteines, but dermcidin's activity spectrum differs. What distinguishes the anionic antimicrobial peptides mechanistically, and is that distinction something the ontology should be able to express?
Suggested experts: antimicrobial peptide biochemists, innate immunity researchers
Experiment: Challenge Gm94 knockout mice and wild-type littermates with a Gram-negative pathogen at a mucosal or cutaneous surface where the gene is expressed - oral or skin inoculation rather than the intraperitoneal route used for the peptide-administration experiment - and measure bacterial burden and survival. The mouse orthologue has the same in vitro activity, so the model is appropriate. This is the single experiment that would convert the two proposed process annotations from sufficiency claims into requirement claims, and it is the gap that no existing experiment addresses.
Hypothesis: Endogenous C5orf46 is required for defence against Gram-negative bacteria at the body surfaces where it is expressed, not merely sufficient to kill them when added.
Type: in vivo loss-of-function infection challenge
Experiment: Quantify the mature peptide in saliva, skin surface washings and plasma by targeted mass spectrometry with a stable-isotope-labelled standard - the small-protein enrichment assay that first detected it in plasma is already suited to the 2-10 kDa window - and compare the concentrations with the published minimum inhibitory concentrations. Without this arithmetic the in vitro activity cannot be known to be physiologically reachable, and the same measurement bears on whether the observed keratinocyte cytotoxicity is relevant at native levels.
Hypothesis: The peptide reaches concentrations in saliva and on skin that are sufficient for the measured bactericidal activity.
Type: targeted quantitative proteomics
Experiment: Distinguish the two alternatives the authors leave open. Measure membrane permeabilisation directly and with time resolution - potassium efflux, propidium iodide or SYTOX uptake, and depolarisation of the inner membrane with a voltage-sensitive dye - against a known pore-former and a known intracellularly-acting antimicrobial as controls, and in parallel track uptake of fluorescently labelled peptide into the cytoplasm. If killing precedes permeabilisation, or if labelled peptide accumulates intracellularly before membrane integrity is lost, the intracellular-target reading is supported. This is the experiment that would license a molecular function term, which is currently withheld for exactly this reason.
Hypothesis: The peptide kills by permeabilising the bacterial membrane, rather than by acting on an intracellular target.
Type: mechanism-of-action biophysics
Experiment: Run side-by-side dose-response curves on the same plate for bacterial killing and for viability of primary human keratinocytes, primary T cells and the lymphoma lines, over the full range from below the minimum inhibitory concentration to 10 micromolar. The published data give bacterial activity from 0.1 micromolar and significant lymphoma killing at 0.4 micromolar, but the two were measured in separate experiments, so the therapeutic index has never actually been determined. The answer decides whether the antibacterial activity is separable from general amphipathic cytotoxicity.
Hypothesis: There is a concentration window in which the peptide kills Gram-negative bacteria without killing host cells.
Type: comparative dose-response cytotoxicity
Experiment: Repeat the siRNA knockdown in A498 and OSRC-2 and attempt rescue with recombinant mature peptide added to the medium, and separately with a re-expressed construct carrying silent siRNA-resistant mutations. If extracellular peptide rescues, the phenotype belongs to the secreted product; if only intracellular re-expression rescues, an intracellular pool is implied; if neither rescues, the phenotype is likely off-target and no GO process term should ever be built on it. This is the cheapest way to decide whether the only loss-of-function result on this gene means anything about its function.
Hypothesis: The renal-carcinoma knockdown phenotype is a function of the secreted peptide.
Type: knockdown-rescue
What is not known β curated, literature-grounded statements of the open unknowns (the inverse of core functions).
Gap: No molecular function is known. The peptide demonstrably kills Gram-negative bacteria, but the route is unresolved: the authors offer cell-envelope damage and an intracellular target as alternatives and settle neither, no molecular target has been identified, and no binding partner has been validated. This is why the review proposes two biological-process terms and no molecular function at all, and why the fourteen protein-binding rows cannot be repaired into something informative - none of them tells us what the peptide does.
OPEN BIOLOGY MF_DARK
What is known: Known: the mature peptide is 64 residues, anionic (pI 4.54), free of cysteines, amphipathic with a central helical hydrophobic segment; it kills four Gram-negative species with measured minimum inhibitory concentrations and concentration dependence over two orders of magnitude; treated bacteria lose shape and disintegrate; Gram-positive pathogens and yeast are unaffected; the growth kinetics resemble those of cell-envelope-interfering agents; and an N-terminal SUMO fusion abolishes the activity, which localises the requirement to a free N-terminus. Unknown: whether the peptide permeabilises the outer or inner membrane or neither; whether it has an intracellular target and enters the cell; what molecule it binds; and why an anionic peptide should engage a negatively charged bacterial surface at all, which is the mechanistically interesting question here and is not addressed anywhere.
Significance: Without it the gene can hold process terms but no molecular function, and the one property that would make it interesting - an anionic antimicrobial peptide acting on Gram-negative bacteria, where the cationic mechanism that explains most human antimicrobial peptides cannot apply - remains unexplained.
Provenance (the field's own admissions):
Gap: Nothing establishes that the endogenous peptide is required for anything. Every functional experiment adds purified peptide from outside: to bacterial cultures, and by intraperitoneal injection into mice already challenged with a pathogen. Those show sufficiency. No knockout, knockdown or patient loss-of-function has been challenged with a pathogen in either human or mouse, so no requirement has been demonstrated for host defence at the skin and mucosal surfaces where the gene is expressed.
OPEN BIOLOGY BP_DARK
What is known: Known: exogenous peptide kills Gram-negative bacteria in vitro and protects mice when administered after challenge; the mouse orthologue does the same, so a genetically tractable model exists; and the gene is expressed most highly in skin, salivary gland and blood vessel, with the protein detectable in plasma and in salivary exosomes. Unknown: whether losing the gene increases susceptibility to Gram-negative infection at any body surface; what concentration the peptide reaches in saliva or on skin relative to its minimum inhibitory concentrations, which is the arithmetic that would decide whether the in vitro activity is physiologically reachable; and whether any of the observed activity operates at endogenous levels.
Significance: The distinction is invisible in GO's evidence codes, so it has to be carried in prose. The two proposed process terms are annotated from a sufficiency result and should be read that way; a knockout challenge in mouse would convert them into requirement claims and is a small experiment relative to that gain.
Provenance (the field's own admissions):
Gap: The gene's functional literature is entirely absent from GO, in every organism, and so is its family's. A QuickGO query by reference returns zero annotations citing PMID:33804835 anywhere in GOA, against 396 for the control reference PMID:19199708. The mouse orthologue Gm94 carries MGI's ND 'no biological data available' root-term annotation in all three aspects, even though the same paper reports that Gm94 itself is bactericidal and protects mice. Bovine Q3T146, the third and last reviewed member of PANTHER PTHR37864, has a single IEA localisation row. So across the whole reviewed family there is not one functional annotation, while a purified-protein activity has been published for two of its three members.
OPEN CURATION BP_DARK
What is known: Known: PANTHER PTHR37864 contains 153 proteins across 426 taxa, of which three are reviewed - human C5orf46 (87 aa), mouse Gm94 Q3V2D2 (93 aa) and bovine Q3T146 (88 aa); the human gene has 16 GOA annotations, 14 of them protein binding, and zero IBA, which is consistent with UniProt's PAN-GO line reporting no evolutionary-model annotations; and the family's only InterPro signature, IPR027950/DUF4576, has no interpro2go mapping, so no term reaches these genes from a domain name either. Unknown: nothing biological - this gap is a curation gap. Curating the mouse row would also open an ISO/ISS route to the human gene that does not currently exist, since transfer requires a source annotation and there is none.
Significance: Two rows on the mouse gene and two on the human gene would replace an explicit 'nothing is known' placeholder with the published activity, and would give the family its first functional annotation. The fix is cheap and it is upstream of this review.
Provenance (the field's own admissions):
Gap: The relationship between the antibacterial activity and the general cytotoxicity is unresolved, and it is the question that decides whether this peptide is a host-defence molecule or simply a membrane-active amphipathic peptide. At 10 micromolar the peptide kills Jurkat and Raji lymphoma lines - but it also kills normal T lymphocytes, HaCaT keratinocytes and mouse embryonic fibroblasts, so it is not tumour-selective and it is not selective for bacteria over host cells at that concentration. Whether there is a concentration window in which it kills bacteria and spares host cells has not been reported.
OPEN BIOLOGY RESIDUAL_SUBGAP
What is known: Known: bacterial killing is measurable from 0.1 micromolar; Jurkat and Raji viability falls significantly at 0.4 micromolar; and normal T cells, keratinocytes and fibroblasts are susceptible at the concentrations tested. Unknown: the dose-response for host-cell killing side by side with the bacterial minimum inhibitory concentrations, which is the only comparison that would establish or exclude a therapeutic window, and whether the keratinocyte sensitivity is compatible with the peptide being made in skin.
Significance: A peptide that kills its own producing tissue at the concentrations that kill bacteria would need a different account of its physiological role, and possibly a different set of GO terms. No cytotoxicity term is proposed in this review precisely because the observation cannot yet be told apart from a general property of an amphipathic peptide.
Provenance (the field's own admissions):
Gap: The one loss-of-function phenotype on record has no mechanism and no connection to the peptide's known activity. siRNA knockdown in two renal-carcinoma lines reduces proliferation and migration and raises apoptosis. That is a cancer-cell-line dependency for a gene whose product is secreted; how a secreted antibacterial peptide would support carcinoma proliferation is not addressed, no rescue was performed, and no molecular function was assayed. It is recorded here rather than converted into a GO process term.
OPEN BIOLOGY BP_DARK
What is known: Known: the gene is reported over-expressed in renal cancer tissue and lines, and three siRNAs reduce proliferation, colony formation and migration and increase apoptosis in A498 and OSRC-2. Unknown: whether the effect is on the secreted peptide or an intracellular pool, whether conditioned medium or recombinant peptide reproduces or rescues it, whether it survives in non-transformed cells, and whether it has anything to do with antibacterial activity. The paper's own title calls it a preliminary study.
Significance: A rescue with recombinant peptide would decide in one experiment whether this is a function of the gene product at all, and would determine whether any GO process term is warranted. Until then, annotating proliferation or apoptosis here would be reading a phenotype as a function.
Provenance (the field's own admissions):
Loading supporting contentβ¦
Download this section (compressed HTML)Loading supporting contentβ¦
Download this section (compressed HTML)Loading supporting contentβ¦
Download this section (compressed HTML)Loading supporting contentβ¦
Download this section (compressed HTML)