The predicted name is a correct domain description and does not assert phosphatase catalysis. The main uncertainty is localization of the exact product: its altered N-terminus carries a computational signal-peptide call, while the characterized MTMR12-MTM1 mechanism acts in the cytoplasm and at muscle membranes.
A0A8I5ZMD5 and Q5FVM6 are same-gene rat Mtmr12 products. Selected residues 41–732 align to reference 57–748 with 691/692 identities; the entire 439-residue myotubularin phosphatase domain is identical. The first 40 selected residues differ from the reference N-terminus. SignalP annotates residues 1–20 as a signal peptide, but this is a prediction rather than demonstrated ER import.
MTMR12 primarily regulates the function of myotubularin protein by affecting protein levels instead of modulating the enzymatic activity
the catalytically inactive MTMR12 only partially but significantly rescues myotubularin function
The complete emitted record is preserved in Mtmr12-protnlm-source.json. Assessments below address the selected protein product; evidence on longer products is identified explicitly.
Myotubularin phosphatase domain-containing protein
CNN (CS 2). The complete myotubularin phosphatase domain is retained, including identity across the reference domain. The wording “domain-containing” does not claim autonomous phosphatase activity and should not be scored as pseudoenzyme overannotation. Sequence comparison.
Membrane
UNC (CS 1). Membrane association is plausible through the normal MTM1 complex and independently through a possible altered trafficking route, but neither demonstrates stable membrane association of this product. A cleavable signal-peptide prediction also does not by itself establish that the mature protein remains membrane bound. PMID:23818870(https://pubmed.ncbi.nlm.nih.gov/23818870/); exact record.
All 1 emitted GO claims are individually assessed in Mtmr12-protnlm-predictions-review.yaml.
The myotubularin family includes active phosphatases and inactive binding partners. MTMR12 belongs to the latter functional class; the full domain’s preservation does not reinstate MTM1-like catalysis. The sequence distinction here concerns the N-terminus and potentially trafficking, not loss of the phosphatase fold.
The primary report has full text and assays zebrafish, mammalian cells and muscle, rather than this alternative rat protein product. The genuine Falcon report provides a useful family and MTM1-interaction synthesis, but its broad localization transfer needs the exact N-terminal caveat. A signal-peptide predictor is insufficient to overturn the established gene-level mechanism or to declare a secreted isoform.
Exact sequence mapping: Mtmr12-bioinformatics/RESULTS.md. Global alignments can place nonhomologous alternative tails opposite gaps or distant residues; only conserved segments and explicitly retained feature intervals support functional transfer.
GOA changes. None for this product: the refreshed GOA still carries only the two UniProt SubCell-mapping IEA rows (GO:0016529 sarcoplasmic reticulum, GO:0030017 sarcomere). No new rows, no retired rows.
Row audit. Both rows remain UNDECIDED. The sarcoplasmic-reticulum row now records a reason: the gene-level localization is well supported (mouse MTMR12, Q80TA6, carries an EXP annotation to GO:0016529 from PMID:23818870; the MTM1-MTMR12 complex "localizes to triads, partly overlapping RyR1-positive sarcoplasmic-reticulum structures but not α-actinin-positive Z-lines" in mouse muscle, per the Falcon report), but whether this alternative-N-terminus product with a predicted signal peptide (SIGNAL 1..20) is targeted the same way is untested. The sarcomere row stays UNDECIDED; the mouse data place the complex at triads rather than Z-lines, so sarcomere is at best a loose description.
core_functions added. One entry: MTM1-binding adaptor (GO:0019902 phosphatase binding) acting in GO:0050821 protein stabilization of MTM1. Rationale: the mechanism is well established at gene level [PMID:23818870 "binds to myotubularin in skeletal muscle"; "MTMR12 primarily regulates the function of myotubularin protein by affecting protein levels instead of modulating the enzymatic activity"], and this rat product retains reference residues 57-748 with 691/692 identity, including the entire myotubularin domain and the C-terminal region (Mtmr12-bioinformatics/RESULTS.md). The core-function description states that the rat product itself has not been assayed.
Not added as NEW rows. Comparator check (QuickGO, 2026-10-04): human MTMR12 (Q9C0I1), mouse (Q80TA6) and the rat reference product (Q5FVM6) carry no GO:0019902 or GO:0050821 annotations; human carries only GO:0005515 IPI rows with MTM1/MTMR2. Proposing NEW annotations on a non-reference TrEMBL product ahead of the reference orthologs is not justified, so the two validator warnings ("core function term not reflected in existing_annotations") are left deliberately.
Open question. Is the alternative N-terminus of A0A8I5ZMD5 a real transcript, and does it change trafficking or MTM1 binding?
ProtNLM2 output and prediction files were not opened or used for this re-review.
Identity. A0A8I5ZMD5 (732 aa, TrEMBL, Ensembl ENSRNOP00000078948) is an alternative gene model of rat Mtmr12. The Swiss-Prot canonical product is Q5FVM6 (748 aa, cDNA BC089876; Q5FVM6 carries no Ensembl xref, so the GRCr8 Ensembl model yields only this variant). A new comparison (Mtmr12-bioinformatics/compare_sequences_rereview.py, RESULTS-rereview.md, sequences fetched live from UniProt) shows: residues 41-732 are 691/692 identical to Q5FVM6 57-748; the first 40 residues (MVVSWVIRKFLCFCCKVCFPFPGYPSFLSFPKPWRISLLP) replace the canonical 1-56 and are not found in mouse Q80TA6 or human Q9C0I1, whose N-termini (MLGKG...GGGGGTKAPKPSF) are conserved with rat Q5FVM6. The hydrophobic segment (max KD 11-mer 1.9 versus 0.3-0.7 in orthologs) is what SignalP calls as a signal peptide. Nothing is missing from the pseudophosphatase domain or C-terminus. The MTM1 CX5R catalytic motif (CSDGWDR) aligns to EENASDL in A0A8I5ZMD5, Q5FVM6 and Q9C0I1, so the catalytic Cys and Arg are absent throughout, as expected for a pseudophosphatase.
Evolutionary grounding. All MTMR12 products (rat, mouse, human, zebrafish) sit in PANTHER PTHR10807:SF37. PAINT (PTHR10807-paint.tsv) propagates cytoplasm (PTN000743854) and membrane (PTN000086845) IBA to the canonical orthologs (QuickGO 2026-10-21: Q5FVM6, Q80TA6, Q9C0I1 all is_active_in GO:0005737 and GO:0016020 IBA). The phosphatase activity GO:0004438 asserted at PTN000086845 is not received by MTMR12 orthologs, consistent with IRD NOT nodes for inactive branches; I read that as a deliberate decision and do not transfer catalytic terms. A0A8I5ZMD5 receives no IBA rows at all (non-canonical entry). The family review interpro/panther/PTHR10807/PTHR10807-review.yaml places this product in SF37, consistent with UniProt; no disagreement on subfamily placement.
Row decisions.
- GO:0016529 sarcoplasmic reticulum (IEA SubCell): UNDECIDED -> ACCEPT. Mouse muscle co-localization with RyR1 and restriction to triads PMID:23818870; mouse Q80TA6 has EXP to this term from that paper. The alternative N-terminus is likely a gene-model feature, and even a used signal peptide would not move the protein out of the SR.
- GO:0030017 sarcomere (IEA SubCell): UNDECIDED -> MODIFY to GO:0030314 junctional membrane complex (triad). The paper says PMID:23818870; GO sarcomere is defined by the filament array between Z discs. MGI curators annotated SR, not sarcomere, from the same paper. Lateral correction.
Core function. Kept GO:0019902 phosphatase binding + GO:0050821 protein stabilization (of MTM1); added locations cytoplasm and sarcoplasmic reticulum, and added PMID:12847286 (human 3-PAP is a catalytically inactive adapter that relocalizes myotubularin to the cytosol). Not GO:0019211 phosphatase activator activity (Reactome TAS on human): PMID:23818870 argues MTMR12 acts on MTM1 protein levels "instead of modulating the enzymatic activity". No NEW rows: per the earlier comparator check, the canonical orthologs carry no GO:0019902/GO:0050821 annotations, and cytoplasm is already IBA on the canonical product; adding them to a non-canonical gene model is not justified, so the "not reflected" validator warnings are deliberate.
Not annotated. Zebrafish morphant phenotypes (triad disorganization, central nuclei, motor defects) are downstream organismal phenotypes and are not process annotations.
Relative to review next. Because A0A8I5ZMD5 is an alternative gene model and the canonical rat Q5FVM6 has no review in the repo, the next relative is Q5FVM6 (same_gene_canonical), folder genes/rat/Mtmr12__Q5FVM6.
Modules. No module under modules/ mentions MTMR12 or myotubularin.