AK4P3

Summary: AMP kinase activity, from the InterPro adenylate-kinase signatures. Appropriate to the sequence; never measured for this protein.

Reason: This row is the InterPro adenylate-kinase signature mapping, which cannot see residue 159. Undecided because the activity is disputed for AK4 itself, and AK4P3 carries the residue blamed for AK4's inactivity. Liu et al. found AK4 inactive as an adenylate kinase in vitro and traced this to a glutamine where other adenylate kinases have a conserved arginine; the Q159R mutation restored activity with GTP [PMID:19130895 "Mutation of the glutamine residue to arginine (Q159R) restored the adenylate kinase activity"]. Noma et al. likewise found no GTP:AMP phosphotransferase activity and no complementation of an E. coli adenylate kinase mutant from overexpressed AK4 [PMID:11485571 "Over expression of AK3, but not AK4"]. Panayiotou et al., by contrast, measured AMP phosphorylation by recombinant AK4 [PMID:19766732 "The enzyme catalyzed the phosphorylation of AMP, dAMP, CMP and dCMP with ATP or GTP as phosphate donors"]. PAINT has taken the inactive reading: the AK4 clade node PTN002764060 carries IRDs negating GO:0004017 and GO:0046899 [file:interpro/panther/PTHR23359/PTHR23359-paint.tsv "PTN002764060"]. AK4P3 keeps Gln159 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Residue 159 (AK4's inactivating Gln) | `Q` in both"], so the sequence predicts AK4's enzymology, whichever reading of it is right. No AK4P3 protein has been assayed. With the template's own activity contested, projection to AK4P3 settles nothing, so neither ACCEPT nor REMOVE is justified.

Summary: Nucleoside diphosphate kinase activity, inherited through the PANTHER node with human AK4 among the donors - the very gene this locus duplicates.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. This row asserts a specific catalytic activity. No AK4P3 protein has ever been assayed, so the assertion rests entirely on 99.1% identity to an enzyme that has been. That is ordinary ISS-grade reasoning and it is not being disputed here, but the caveat belongs on the row: the activity is inferred, not measured, for this gene product. The donor here is P27144 itself, which holds an IDA for this term, so the node placement is sound and the transfer is to a near-identical sequence.

Summary: Nucleoside diphosphate kinase activity from the EC 2.7.4.6 mapping carried on the UniProt entry.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. This row asserts a specific catalytic activity. No AK4P3 protein has ever been assayed, so the assertion rests entirely on 99.1% identity to an enzyme that has been. That is ordinary ISS-grade reasoning and it is not being disputed here, but the caveat belongs on the row: the activity is inferred, not measured, for this gene product. The EC number is assigned to this entry by HAMAP-Rule MF_03170 rather than measured, so this row and the IBA row above share a single underlying inference despite different evidence codes.

Summary: ATP binding. The best-supported molecular function row: the Walker A P-loop is present and unaltered.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. Unlike the catalytic rows this one needs no assay to be credible - the binding determinant is a conserved structural motif and it is intact, at the same position as in AK4.

Summary: Cytoplasm, inherited from a very deep pan-adenylate-kinase node with 37 donors spanning plants, yeast, fly, fish and mammals.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. Worth separating from the other localization rows: it comes from a much deeper node (PTN000599576) than the mitochondrial matrix row (PTN008598552), and most adenylate kinases in that node are cytosolic. It is not wrong - mitochondrion is part of cytoplasm in GO - but it is the least informative location statement on the gene, and it does not corroborate the matrix rows.

Summary: Mitochondrion, from the InterPro AK3/AK4 signatures. The general parent of the matrix term.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. A parent of the matrix rows and unobjectionable if the protein exists.

Summary: Mitochondrial matrix, from the AK3/AK4 subfamily node whose donors include AK4 itself, which has EXP evidence for this location.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. Localization is better placed than the catalytic claims: the mitochondrial targeting information is part of the sequence, and the sequence is 99.1% identical, so if the protein is made it should go where AK4 goes. Still untested for this gene product.

Summary: Mitochondrial matrix from the UniProt SubCell vocabulary and UniRule - the same inference as the IBA row, not a second observation.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. Shares its basis with the IBA row above; both derive from AK4's localization.

Summary: ADP biosynthetic process, from UniRule.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. This is a process term that presupposes the catalysis asserted by the molecular function rows, so it inherits their caveat rather than adding independent support: no AK4P3 protein has been assayed, and the flux implied here has never been measured for this gene product.

Summary: Nucleoside triphosphate biosynthetic process, from the AK3/AK4 subfamily node.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. This is a process term that presupposes the catalysis asserted by the molecular function rows, so it inherits their caveat rather than adding independent support: no AK4P3 protein has been assayed, and the flux implied here has never been measured for this gene product.

Summary: Phosphotransferase activity, phosphate group as acceptor. A broad parent, well supported by the conserved nucleotide-binding fold.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. A broad parent claim supported by the fold itself rather than by any specific reaction, so it stands independently of whether the precise substrate specificity transfers.

Summary: Nucleobase-containing compound kinase activity - the informative parent of the specific kinase terms, following from the intact P-loop.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. This is a parent term rather than a specific activity claim, and the structural basis for it is directly visible in the sequence, so it is the least exposed of the molecular function rows.

Summary: AMP metabolic process, from the PANTHER node with AK4 among the donors.

Reason: AMP metabolic process follows from the AMP kinase activity alone, so it stands or falls with GO:0004017. The IBA comes from node PTN008598552, above the AK4 clade node that carries the IRDs. Undecided because the activity is disputed for AK4 itself, and AK4P3 carries the residue blamed for AK4's inactivity. Liu et al. found AK4 inactive as an adenylate kinase in vitro and traced this to a glutamine where other adenylate kinases have a conserved arginine; the Q159R mutation restored activity with GTP [PMID:19130895 "Mutation of the glutamine residue to arginine (Q159R) restored the adenylate kinase activity"]. Noma et al. likewise found no GTP:AMP phosphotransferase activity and no complementation of an E. coli adenylate kinase mutant from overexpressed AK4 [PMID:11485571 "Over expression of AK3, but not AK4"]. Panayiotou et al., by contrast, measured AMP phosphorylation by recombinant AK4 [PMID:19766732 "The enzyme catalyzed the phosphorylation of AMP, dAMP, CMP and dCMP with ATP or GTP as phosphate donors"]. PAINT has taken the inactive reading: the AK4 clade node PTN002764060 carries IRDs negating GO:0004017 and GO:0046899 [file:interpro/panther/PTHR23359/PTHR23359-paint.tsv "PTN002764060"]. AK4P3 keeps Gln159 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Residue 159 (AK4's inactivating Gln) | `Q` in both"], so the sequence predicts AK4's enzymology, whichever reading of it is right. No AK4P3 protein has been assayed. With the template's own activity contested, projection to AK4P3 settles nothing, so neither ACCEPT nor REMOVE is justified.

Summary: AMP metabolic process, from UniRule. Same claim as the IBA row and the same underlying inference.

Reason: AMP metabolic process follows from the AMP kinase activity alone, so it stands or falls with GO:0004017. Undecided because the activity is disputed for AK4 itself, and AK4P3 carries the residue blamed for AK4's inactivity. Liu et al. found AK4 inactive as an adenylate kinase in vitro and traced this to a glutamine where other adenylate kinases have a conserved arginine; the Q159R mutation restored activity with GTP [PMID:19130895 "Mutation of the glutamine residue to arginine (Q159R) restored the adenylate kinase activity"]. Noma et al. likewise found no GTP:AMP phosphotransferase activity and no complementation of an E. coli adenylate kinase mutant from overexpressed AK4 [PMID:11485571 "Over expression of AK3, but not AK4"]. Panayiotou et al., by contrast, measured AMP phosphorylation by recombinant AK4 [PMID:19766732 "The enzyme catalyzed the phosphorylation of AMP, dAMP, CMP and dCMP with ATP or GTP as phosphate donors"]. PAINT has taken the inactive reading: the AK4 clade node PTN002764060 carries IRDs negating GO:0004017 and GO:0046899 [file:interpro/panther/PTHR23359/PTHR23359-paint.tsv "PTN002764060"]. AK4P3 keeps Gln159 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Residue 159 (AK4's inactivating Gln) | `Q` in both"], so the sequence predicts AK4's enzymology, whichever reading of it is right. No AK4P3 protein has been assayed. With the template's own activity contested, projection to AK4P3 settles nothing, so neither ACCEPT nor REMOVE is justified.

Summary: ATP metabolic process, from UniRule UR000166132.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. This is a process term that presupposes the catalysis asserted by the molecular function rows, so it inherits their caveat rather than adding independent support: no AK4P3 protein has been assayed, and the flux implied here has never been measured for this gene product.

Summary: GTP metabolic process, from UniRule. AK4's distinctive feature is its GTP preference, so this is the term most specific to the AK4 subfamily.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. This is a process term that presupposes the catalysis asserted by the molecular function rows, so it inherits their caveat rather than adding independent support: no AK4P3 protein has been assayed, and the flux implied here has never been measured for this gene product. Worth noting that GTP usage is what distinguishes AK4 from the cytosolic adenylate kinases, so if this locus does make protein, this is among the more discriminating predictions to test.

Summary: Nucleoside triphosphate adenylate kinase activity (EC 2.7.4.10) - the activity AK4 is best known for.

Reason: This row is the EC 2.7.4.10 mapping, the GTP/ATP:AMP reaction Q159R restores in AK4. Undecided because the activity is disputed for AK4 itself, and AK4P3 carries the residue blamed for AK4's inactivity. Liu et al. found AK4 inactive as an adenylate kinase in vitro and traced this to a glutamine where other adenylate kinases have a conserved arginine; the Q159R mutation restored activity with GTP [PMID:19130895 "Mutation of the glutamine residue to arginine (Q159R) restored the adenylate kinase activity"]. Noma et al. likewise found no GTP:AMP phosphotransferase activity and no complementation of an E. coli adenylate kinase mutant from overexpressed AK4 [PMID:11485571 "Over expression of AK3, but not AK4"]. Panayiotou et al., by contrast, measured AMP phosphorylation by recombinant AK4 [PMID:19766732 "The enzyme catalyzed the phosphorylation of AMP, dAMP, CMP and dCMP with ATP or GTP as phosphate donors"]. PAINT has taken the inactive reading: the AK4 clade node PTN002764060 carries IRDs negating GO:0004017 and GO:0046899 [file:interpro/panther/PTHR23359/PTHR23359-paint.tsv "PTN002764060"]. AK4P3 keeps Gln159 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Residue 159 (AK4's inactivating Gln) | `Q` in both"], so the sequence predicts AK4's enzymology, whichever reading of it is right. No AK4P3 protein has been assayed. With the template's own activity contested, projection to AK4P3 settles nothing, so neither ACCEPT nor REMOVE is justified.

Summary: Nucleoside monophosphate kinase activity from the EC 2.7.4.4 mapping. Genuine AK4 carries this term too.

Reason: GO:0050145 is the direct parent of GO:0004017 and follows the same evidence: the nucleoside diphosphate kinase activity accepted on other rows does not satisfy it, because that enzyme acts on diphosphates, not monophosphates. Undecided because the activity is disputed for AK4 itself, and AK4P3 carries the residue blamed for AK4's inactivity. Liu et al. found AK4 inactive as an adenylate kinase in vitro and traced this to a glutamine where other adenylate kinases have a conserved arginine; the Q159R mutation restored activity with GTP [PMID:19130895 "Mutation of the glutamine residue to arginine (Q159R) restored the adenylate kinase activity"]. Noma et al. likewise found no GTP:AMP phosphotransferase activity and no complementation of an E. coli adenylate kinase mutant from overexpressed AK4 [PMID:11485571 "Over expression of AK3, but not AK4"]. Panayiotou et al., by contrast, measured AMP phosphorylation by recombinant AK4 [PMID:19766732 "The enzyme catalyzed the phosphorylation of AMP, dAMP, CMP and dCMP with ATP or GTP as phosphate donors"]. PAINT has taken the inactive reading: the AK4 clade node PTN002764060 carries IRDs negating GO:0004017 and GO:0046899 [file:interpro/panther/PTHR23359/PTHR23359-paint.tsv "PTN002764060"]. AK4P3 keeps Gln159 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Residue 159 (AK4's inactivating Gln) | `Q` in both"], so the sequence predicts AK4's enzymology, whichever reading of it is right. No AK4P3 protein has been assayed. With the template's own activity contested, projection to AK4P3 settles nothing, so neither ACCEPT nor REMOVE is justified.

Summary: Nucleobase-containing small molecule metabolic process - a broad parent that follows from family membership alone.

Reason: Context for the whole gene. HGNC names this locus 'adenylate kinase 4 pseudogene 3' with locus_type pseudogene, while Ensembl gives the same locus biotype protein_coding and UniProt lists it unreviewed at [file:genes/human/AK4P3/AK4P3-uniprot.txt "PE 1: Evidence at protein level;"]. All 19 of its GO annotations are IEA or IBA, applied by automatic pipelines - chiefly HAMAP-Rule MF_03170, InterPro2GO and PANTHER - on similarity to genuine human AK4 (P27144), which is itself a well-characterised enzyme carrying IDA and EXP evidence for these activities. So the question is not whether the terms describe AK4: they do. It is whether they should be asserted of this locus. Two lines settle more of that than one might expect. A sequence analysis run for this review finds the ORF near-identical to AK4 [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Identity to AK4 | **221/223 (99.1%)** |"] with the catalytic machinery intact [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "| Catalytic P-loop (Walker A) | `GPPGSGK` at position 12 — **intact in both** |"], so [file:genes/human/AK4P3/AK4P3-bioinformatics/RESULTS.md "The sequence gives no evidence of pseudogenisation at the protein level."] And the locus is genuinely transcribed and its ORF is translation-competent: long-read sequencing recovered the spliced transcript from human RNA [PMID:33971925 "we amplified the 5′ exons and coding sequence of four spliced pseudogenes with intact ORFs (HMGB1P1, AK4P3, IFITM3P2 and RPL13AP20) and cloned them into a vector with a C-terminus 3XHA tag"] and [PMID:33971925 "Transfection into HEK293T cells resulted in clear translation of the HMGB1P1 and AK4P3 transcripts"]. What is NOT shown anywhere is endogenous protein production or any assay of this protein's activity [file:genes/human/AK4P3/AK4P3-deep-research-falcon.md "It does **not** establish endogenous protein production, enzymatic activity, substrate specificity, mitochondrial localization, pathway membership, disease causality, or physiological function."]. Broad enough that it holds on family membership without committing to a specific reaction.