ACTL10: is the Swiss-Prot sequence the protein, and what do its two IBA rows rest on?

Generated by uv run python analyze_actl10.py. Every number is computed at run time
from the UniProt and Ensembl REST APIs, RCSB coordinate files, QuickGO, and files
committed in this repository. Nothing is hard-coded from a previous run.

1. ACTL10 orthologue lengths do not follow the phylogeny

UniProt holds 87 mammalian entries whose gene name is exactly ACTL10, across 84 organisms. Lengths run from 169 to 487 aa. The human entry is 245 aa, shared with 25 other entries; the modal length is 245 aa (26 entries). 59 entries are longer than human, of which 50 are 340 aa or more - i.e. long enough to span the actin fold that the human entry does not.

If the variation were biological it would track the species tree. It does not: in 4 of 4 sister-taxon pairs tested, two members of the same family carry different lengths.

Organisms carrying more than one ACTL10 length in UniProt at once (same species, different entries):

• Ovis aries: [335, 344, 379]

Length histogram:

2. Q5JWF8 begins mid-fold: the human genome encodes the missing region in frame

Q5JWF8 (ACL10_HUMAN, UniProtKB reviewed (Swiss-Prot)) is 245 aa. Its MANE transcript ENST00000677665 is a single exon (1 exon, chr20:33666943-33668525) with the CDS at 33667498-33668235. The annotated 5' leader is therefore 555 nt of the same exon, contiguous with the CDS.

Translating that leader in the CDS reading frame (frame offset 0; the CDS translation was first asserted to equal the Swiss-Prot sequence, so the frame is proven, not assumed) gives 185 codons containing 1 in-frame stop. After the last stop there are 167 uninterrupted codons running straight into the annotated initiator:

DRGREVSGRRSRVARRSGAFSPAAGPSPKVASVSGSRRVHRPSSLGRIAVVVDQGSGFTK
AGFAGENQPRIVLKSSSLVPSWDRPVLPGAPGCELAGGVARAHPIKHGVVADWEALEGLW
ERLLVGGLRVCPEQWPVLVSDSPLAPPAGRERVAELLFETLAVPACH

In-frame ATG codons inside that open stretch: none.

That translation is actin. Aligned globally to human beta-actin the extended (412 aa) form scores 553.0 at 34.7% identity, against 244.0 at 33.9% for the 245-aa Swiss-Prot sequence alone.

Against the long ACTL10 orthologues the gain is decisive:

Where the ancestral initiator went

• Sapajus apella (A0A6J3JAC8, 368 aa): its Met1 aligns to position 45 of the extended human ORF, where the human genomic codon is CTG (chr20:33667129), encoding L - ATG: False.
• Callithrix jacchus (A0A5F4WJE9, 368 aa): its Met1 aligns to position 45 of the extended human ORF, where the human genomic codon is CTG (chr20:33667129), encoding L - ATG: False.

So the human locus has lost the initiator codon its orthologues use, and the next in-frame ATG lies ~120 codons downstream, which is where Swiss-Prot, RefSeq and MANE all begin the protein. The intervening actin-homologous coding sequence is still present, in frame, and free of stop codons.

What this does and does not establish. It establishes that the sequence in Q5JWF8 is not the whole of ACTL10's actin homology, and therefore that any residue tally computed from Q5JWF8 measures the annotation boundary as much as the protein. It does not establish which product the human cell makes: a lost initiator with a conserved downstream reading frame is compatible both with a genuinely N-terminally shortened human protein and with initiation at a non-ATG codon or an unannotated upstream exon. That question needs N-terminal proteomics, not sequence analysis.

3. Actin's nucleotide site and protomer interface, with absence separated from substitution

Nucleotide-site contacts come from PDB 2BTF chain A (ligands ATP, SR, 4.0 A heavy-atom cutoff): 19 residues. Filament-interface contacts come from PDB 6DJO chain C of 4 protomers (4.0 A): 38 residues. Both are actin-only assemblies, and both match the structures and cutoffs used by the committed ACTL8 analysis, so the columns are comparable by construction.

outside span counts contact positions the query sequence does not reach at all - an absence in the annotation. internal gap is a deletion inside the aligned span. Only the remaining columns are substitutions.

Filament protomer interface

Nucleotide site

Per-residue, and why the truncation matters most here

Short column names: ACTB = human beta-actin; ext = human ACTL10 extended ORF; 245 = Q5JWF8 as annotated; Sap = Sapajus ACTL10 (368 aa); mus = mouse Actl10 (346 aa); L8 = ACTL8.

~ conservative, ** non-conservative, ABSENT the query does not reach this position.

The positions Q5JWF8 fails to reach are exactly 13, 14, 15, 16, 18 - actin's phosphate-binding loop 1 (the DNGSGMCK motif that grips the nucleotide beta-phosphate, and the most diagnostic single feature of the actin fold). In the extended human ORF those same positions read G:identical, S:identical, G:identical, F:conservative, K:identical, i.e. the loop is intact. So the one part of the nucleotide site that the committed ACTL8 panel scored as missing from ACTL10 is the part that is present in the genome and merely absent from the annotation. Note that the mouse 346-aa entry also begins downstream of this loop, so its scores at these positions are alignment-edge artefacts rather than substitutions.

Cross-check against the committed ACTL8 analysis

This script must reproduce ACTL8's published filament-interface tallies on the shared rows, where its single gap column equals internal gap + outside span here. The run aborts if it does not.

Alignment sensitivity

4. What ACTL10's two IBA rows rest on

GO:0015629 actin cytoskeleton (cellular_component, IBA, is_active_in)

Reference GO_REF:0000033; 25 WITH/FROM tokens, of which 24 are protein identifiers. Resolved: 24/24. Carrying their own experimental-code annotation for this term or a descendant: 24. Ambiguous lookups (>1 UniProt hit): 6. Resolved only to an unreviewed entry: 2.

GO:0005200 structural constituent of cytoskeleton (molecular_function, IBA, enables)

Reference GO_REF:0000033; 11 WITH/FROM tokens, of which 10 are protein identifiers. Resolved: 10/10. Carrying their own experimental-code annotation for this term or a descendant: 10. Ambiguous lookups (>1 UniProt hit): 2. Resolved only to an unreviewed entry: 0.

GO:0007010 cytoskeleton organization (biological_process, IEA, involved_in)

Reference GO_REF:0000108; 1 WITH/FROM tokens, of which 0 are protein identifiers. Resolved: 0/0. Carrying their own experimental-code annotation for this term or a descendant: 0. Ambiguous lookups (>1 UniProt hit): 0. Resolved only to an unreviewed entry: 0.

GO:0005200 by PANTHER node, all human IBA annotations

QuickGO returns 43 human GO:0005200 IBA annotations in total. Grouped by the donating PANTHER node:

PAINT's own verdict on GO:0005200 inside PTHR11937

The cached PAINT export for PTHR11937 has 86 rows. GO:0005200 is asserted by IBD at:

and then negated on descent at 8 nodes: