The paper that both legacy TAS annotations on ACTL7B trace to. It reports cDNA selection, genomic sequencing, chromosomal mapping, Northern-blot tissue expression and mutation screening in familial dysautonomia patients. It contains no protein-level experiment, no localisation study and no cytoskeletal assay, so it cannot support either "structural constituent of cytoskeleton" or "actin cytoskeleton".

Establishes the testis-restricted expression of ACTL7B, in contrast to the broadly expressed paralogue ACTL7A.

Original mouse immunolocalisation, and the source of the mouse Actl7b GO:0005737 IDA that anchors the cytoplasm IBA on human ACTL7B. T-ACTIN 1 is mouse Actl7b and T-ACTIN 2 is mouse Actl7a; the two paralogues were placed in different compartments.

The authors already noted that the T-ACTINs retain only part of actin's conserved architecture, consistent with the residue-level audit performed for this review.

Human sequencing of 282 sterile and 89 fertile men found polymorphisms in both genes but established no causal link, so there is still no human genetic evidence assigning a function to ACTL7B.

The source of all four GO:0005515 protein binding rows on ACTL7B. It is a systematic yeast two-hybrid matrix built around neurodegeneration-related baits, generating tens of thousands of candidate interactions; ACTL7B, a testis-restricted spermatid protein, appears among the prey. No orthogonal validation of any ACTL7B interaction is reported.

Mouse Actl7b knockout. Establishes the subcellular distribution of ACTL7B during spermiogenesis, including the developing acrosome and the nucleus of early spermatids.

The knockout phenotype is infertility with oligoteratozoospermia and multiple morphological abnormalities of the flagellum and sperm head.

States explicitly that the testis-specific ARPs, including ACTL7B, were functionally uncharacterised before this work, which is why the pre-existing GO annotations rest on fold-level inference rather than data.

Independent mouse Actl7b-null line confirming the infertility and spermatid morphogenesis phenotype, with a defined onset at spermiogenesis step 9.

Identifies DYNLL1 and DYNLL2 as ACTL7B partners by co-immunoprecipitation from wild-type testis, that is, at endogenous levels in the native tissue.

Reciprocal co-immunoprecipitation confirms the interaction in the opposite direction.

Loss of ACTL7B changes where DYNLL1 and DYNLL2 are, without changing how much of them there is, so this is a localisation phenotype rather than a stability phenotype.

Ectopic expression of ACTL7B in HEK cells redistributes DYNLL1 and DYNLL2, providing a gain-of-function counterpart to the knockout result.

Records the nuclear pool of ACTL7B alongside the dominant cytoplasmic pool, refining the cytoplasm-only assignment of PMID:12672658.

Confirms an intranuclear pool of ACTL7B in mouse spermatocytes and round spermatids and reports a conserved putative nuclear localisation sequence.

Reports that intranuclear HDAC1 and HDAC3 localisation is lost in the absence of ACTL7A or ACTL7B.

The proposed mechanism, ACTL7B substituting for conventional nuclear ARPs on the HSA domains of INO80 and SWI/SNF remodellers, is explicitly a computational docking result rather than a binding measurement, and is treated here as a hypothesis.

PAINT explicitly rejected GO:0005200 for the clade containing ACTL7B. Node PTN008986528 carries GO:0005198 by IBA and GO:0005200 as an IRD with negated=true, both inherited from node PTN000940351, whose GO:0005200 IBD is seeded from conventional actins and Arp2/Arp3.

The protomer-protomer interface of the F-actin filament is not conserved in ACTL7B. Of 74 contact positions computed from the middle protomer of 8A2S, 26 are identical and 34 differ non-conservatively (35.1% identity), versus 97-100% for conventional actins.

The largest single interface segment, actin residues 38-45 (the DNase-I-binding D-loop, the principal longitudinal inter-protomer contact), is entirely replaced in ACTL7B. PRHQGVMV becomes RCPEAADA, 0 of 8 identical, against 8 of 8 in ACTB.

By contrast the nucleotide pocket is only partly degenerate. ACTL7B retains the P-loop serine and lysine (S14, K18) and the catalytic Gln137, while losing G15, D157, R210, T303 and Y306. Nucleotide binding is therefore untested rather than excluded, and no ATP-binding annotation is being contested.

Every protein-level WITH/FROM source of every IBA row carries its own experimental evidence for the propagated term, including yeast ARP9, which has IDA and IPI for nucleus. No row in this review is criticised on the grounds that its source was guessing.

UniProt makes no functional claim for ACTL7B beyond family membership, and its only subcellular-location statement is an inference (ECO:0000250, By similarity) rather than an observation, which is the ultimate source of the GO:0005856 IEA.

The entry asserts only family membership, with ECO:0000305 (curator inference).

The protein is detected at protein level, so the gap is functional and biochemical characterisation of the human protein rather than evidence of its existence.

Conserved-domain databases already treat the ACTL7 clade as distinct from generic actin, assigning a clade-specific nucleotide-binding-domain model.

The affinage record's own GO grounding proposes cytoskeletal protein binding as ACTL7B's molecular activity. No measurement supports it: no actin-binding or cytoskeletal-protein-binding assay has been reported for ACTL7B, and the one documented interaction is with dynein light chains. Recorded here because it is the same fold-to-activity leap that produced the legacy TAS annotations.