ADIRF (Q15847) bioinformatics results

Generated by analyze_adirf.py from live NCBI Gene, UniProt and QuickGO
queries. Regenerate with uv run python analyze_adirf.py.

A. ADIRF is absent from the mouse/rat lineage

UniProt entries with gene name ADIRF: mouse 0, rat 0, human 3.

All positive controls non-zero: True — so the zeros are real absences and not rejected queries.
ADIRF absent from Muroidea: True; retained in the sister rodent clade Sciuridae: True (9 genes). The loss is therefore localised to Muroidea rather than being a rodent-wide or annotation-wide absence.

Consequence for the GO record: every functional experiment on ADIRF was performed by ectopic expression in mouse 3T3-L1 preadipocytes, a cell line from a lineage that has no ADIRF gene. Murine adipogenesis therefore proceeds without any ADIRF orthologue.

B. Where the two interpro2go terms from IPR034450 land

withFrom=InterPro:IPR034450 returns 1512 annotations over 789 distinct gene products, supplying exactly the two terms interpro2go maps: GO:0005634, GO:0045600.

Clade distribution of GO:0045600 recipients:

GO:0045600 carries only_in_taxon NCBITaxon:6072 (Eumetazoa). Recipients outside Metazoa altogether: 0. Recipients that are both non-vertebrate and larger than 200 aa: 217.

Examples of non-vertebrate oversized recipients of the adipocyte term:

C. The family signature is promiscuous on low-complexity sequence

PTHR39227 has 768 members in UniProtKB. Length distribution:

Human ADIRF is 76 aa with its three commonest residues accounting for 43.4% of the sequence (A 18.4%, Q 14.5%, K 10.5%).

Local alignment of human ADIRF against genuine orthologues (positive controls) and against oversized family members that receive the GO terms:

The discriminator used is alignment coverage of the 76-aa query, not identity. Every genuine orthologue aligns over at least 68 of the 76 residues; every oversized member aligns over fewer.

Identity is reported descriptively only. Orthologue identity range 43.4-100.0%; oversized-member range 6.6-19.7%; the ranges do not overlap (True), separated by 23.7 points. But identity is the weaker instrument: the largest gap anywhere in the pooled identity distribution is 35.5 points, between 50.0% and 85.5%, which falls inside the genuine orthologues (the bird/mammal split) rather than between orthologues and spurious matches. An identity cut placed at the largest observed gap would therefore have misclassified chicken and pigeon ADIRF. That is why no identity threshold is used or derived here.

D. HPA calls two main locations; GOA imported one

HPA record for ADIRF (ENSG00000148671), IF reliability Supported, main subcellular locations Nucleoplasm, Cytosol.

GOA rows attributed to the HPA immunofluorescence route (GO_REF:0000052): GO:0005654.

Terms expected from HPA's main locations: GO:0005654, GO:0005829. Missing from ADIRF's GOA record entirely: GO:0005829.

Positive control: GAPDH (P04406) is also called Plasma membrane, Cytosol by HPA and does carry GO:0005829 in GOA (term GO:0005829, resolved through the same mapping the subject uses: True) — so the missing ADIRF row is a real gap in the import, not a broken query or a wrong term id.

E. Reach of PANTHER:PTN008674116, the node behind both IBA rows

PANTHER:PTN008674116 carries 14 IBA annotations to 7 gene products: GO:0005634 (7 recipients), GO:0045600 (7 recipients).

IBD seeds named in the WITH/FROM, with the db field read rather than flattened away: UniProtKB:Q15847 x14. The seed is the gene under review, so both IBA rows are self-referential: True. Human ADIRF is the only member of the family with any experimental annotation, which is why it is the sole seed.

Every recipient checked individually against the same coverage criterion used in section C (at least 68 of the 76 query residues aligned) rather than assumed homogeneous:

6 of 7 recipients are genuine orthologues; 1 is not. This was not the expected result -- the node is small and curated, and the first version of this review asserted that all its recipients were 71-76 aa orthologues. The guard above caught that as false.

• A0A5F8H3S4 (Monodelphis domestica (Gray short-tailed opossum), 447 aa, "Uncharacterized protein") aligns over only 62 of 76 residues at 15.8% identity. It is a tandem-repeat protein: repeat period 22 at 85.4% periodicity, and its three commonest residues are 52.1% of the sequence. It nonetheless receives both of human ADIRF's IBA terms, including is_active_in nucleus.

Reciprocal half — does a real ADIRF gene exist in that organism, and did it get anything?

So the organism has a real ADIRF gene, but UniProt's proteome for it contains no ADIRF-sized member of the family — only the tandem-repeat protein. The root cause is therefore upstream of PAINT: the tree was given the wrong protein for that species and annotated the sequence it had. PAINT's placement of the six genuine orthologues is correct.

Muroid recipients: none — mouse and rat are absent from this node's reach because Muroidea have no ADIRF gene (section A), not because PAINT declined to annotate them.

F. Teleost conservation, adjudicated by alignment rather than by symbol count

The 103 Actinopterygii figure in section A is an NCBI symbol/alias count -- orthology already asserted by an annotation pipeline -- so it cannot settle conservation, and the only teleost sequence section C aligns is a 938-aa UniProt family member that lands in the spurious bin. This section settles it on sequence.

UniProt's IPR034450 family has 50 teleost members and 0 of them are ADIRF-sized. That is why no UniProt-based query could answer this: the family's teleost content is entirely oversized matches. The real teleost ADIRF proteins are annotated in RefSeq.

Aligned under the same criterion as sections C and E (at least 68 of the 76 query residues):

Composition control. 1 of 30 composition-matched shuffles of the Danio sequence (identical amino-acid content, order destroyed, deterministic seeds) pass the coverage criterion, and their identity range is 7.9-27.6% against the real proteins' 38.2%. So the criterion is not satisfied by amino-acid composition alone -- the obvious risk for an Ala/Gln-rich 76-aa protein -- though the non-zero shuffle pass rate is why the identity margin is reported alongside coverage rather than coverage being taken as sufficient on its own.

Result: teleost orthology supported by alignment = True. Note that coverage and identity disagree in direction here: both fish proteins align over 71 of 76 residues, comfortably past the criterion, yet at 38.2% identity they sit below the orthologue identity floor established by the birds in section C (43.4%). Coverage is the criterion this analysis committed to before the fish were examined, and greater divergence is expected across a longer branch, but the disagreement is recorded rather than resolved by picking the instrument that gives the wanted answer.

What these results do and do not support

• They do not bear on whether human ADIRF promotes adipogenesis. The human GO:0045600 IDA rests on its own experiment and is untouched here.
• Section A bounds the interpretation of that experiment: it was a gain-of-function in a lineage lacking the gene, so it cannot establish that ADIRF is normally required for adipogenesis.
• Sections B and C are about the IPR034450 propagation route, not about human ADIRF: the signature matches unrelated low-complexity proteins and carries a vertebrate-specific process term to them.
• No claim is made that the genuine non-mammalian orthologues should lose the term; the measurement separates orthologues from spurious matches and says nothing about which is biologically right for a fish or bird.