GAMT methylates guanidinoacetate using S-adenosylmethionine (SAM), producing creatine and S-adenosylhomocysteine (SAH). The strongest selected human evidence is the liver activity experiment in PMID:8651275(https://pmc.ncbi.nlm.nih.gov/articles/PMC1914613/): scanned original pages 915–917 include radiomethyl transfer, chromatographic identification of creatine, substrate-free blanks and reference-enzyme controls. The patient deficiency observations complement this positive reaction assay. They are not the sole basis for assigning catalytic activity or creatine biosynthesis.
The normal publication cache for that paper is abstract-only. The original scanned pages were read separately; the endpoint labeled as a PDF returned HTML and is excluded from the original-page evidence. PMID:9325156(https://pubmed.ncbi.nlm.nih.gov/9325156/) supplies an authentic complete abstract describing the human reaction and gene mapping, with full Methods and figures unavailable. Its negative result for a coding mutation in jittery mice does not contradict human GAMT activity.
Reactome R-HSA-71286 explicitly connects SAM and guanidinoacetate inputs to SAH, creatine and a proton. Catalyst R-HSA-71276 resolves to human Q14353 and cytosol. The official HTML reaction and participant records were inspected; failed ContentService requests are not represented as successful JSON evidence. Creatine metabolism includes downstream transport, phosphorylation and creatinine formation, but GAMT performs the biosynthetic methylation step. Broad methyltransferase and creatine-metabolism annotations are therefore refined to the verified specific descendants.
The inherited core-only SAM-catabolism term describes another aspect of this same reaction. A current exact-term comparator search found experimental rat Gamt support and mouse Gamt, Gnmt and Pcmt1 annotations; the term is used for enzymes performing SAM-dependent methyl transfer. The verified typed ancestors do not make it redundant with creatine biosynthesis. This retains one core with two process facets, without adding a NEW annotation or a quantitative methyl-donor-homeostasis claim. The cached GO-CAM index has no Q14353/GAMT match; this is a dated index observation, not a statement about every live model.
The actual PTHR32379 tree places the exact human 236-residue target at AN17/PTN002529431. The catalytic and creatine-biosynthesis assertions arise at PTN001285367; nuclear and cytoplasmic assertions arise at PTN001282244. All four asserted nodes lie on the target path, and literal IBD donors agree with the current incoming and retained main PAINT records. Human self-donor evidence is legitimate. Experimental fungal localization records support the ancestral compartment assertions; no target-specific localization loss was established. The selected yeast Rmt2 abstract describes nuclear and cytoplasmic localization, while its full original panels and the high-throughput donor images were not independently inspected. Human cytosolic evidence is nonexclusive; it does not disprove a nuclear pool or establish that pool's biochemical role. The legitimate nuclear IBA is retained as non-core, with NO_FAILURE_NON_CORE propagation metadata; substrate divergence alone is not a localization-loss argument. The broad cytoplasm row is refined to cytosol using the direct typed part_of relationship and the resolved human catalyst, while its legitimate ancestral placement is retained.
UniProt Q14353 declares two natural products. The displayed Q14353-1 sequence has 236 residues. Q14353-2 has 269 residues and matches the direct FASTA and exact VSP_042722 replacement of canonical residues 191–236. The published UniProt record maps Q14353-2 to NM_138924.3/NP_620279.1. PMID:26003046(https://pubmed.ncbi.nlm.nih.gov/26003046/) reports no measurable activity for wild-type transcript 2; the official NCBI Gene 2593 variant description also reports no detected GFP-fusion protein under the study conditions. This is an expression-confounded result, not a purified-protein catalytic-inactivity test. Original Methods were not accessed. Equal activity, stability or localization of the two products is not established. The processed chain 2–236 and patient disease transcripts are separate from natural splice-product identity. Complete current sequences remain in the compact evidence record.
BioPlex2 and BioPlex3 293T tables contain the GAMT–FARS2 pair, at original lines 19992 and 63913 respectively. Tagged-protein affinity-purification mass spectrometry supports co-association without establishing direct binary binding or a defined GAMT regulatory role. Absence of that edge from the inspected HCT116 network is not a universal biological negative. The source methods and selected text are from PMID:28514442(https://pubmed.ncbi.nlm.nih.gov/28514442/) and PMID:33961781(https://pubmed.ncbi.nlm.nih.gov/33961781/).
The retained HuRI PSI dataset for PMID:32296183(https://pubmed.ncbi.nlm.nih.gov/32296183/) contains Q14353-1–Q969Q5 at lines 10815, 31161 and 103319, and Q14353-1–Q9HCM9-2 at lines 2887, 61333, 77131, 149193, 165485 and 170884. These are array, prey-pooling and validated two-hybrid records from human ORFs in yeast. Nine records do not mean nine independent biological interactions. The exact target clone, partner suffix and historical publication placeholder remain in provenance. These positive observations do not define a more informative GAMT molecular activity, so the four bare protein-binding annotations are removed as uninformative, without denying the associations.
The muscle-contraction TAS remains UNDECIDED. PMID:8547310(https://pubmed.ncbi.nlm.nih.gov/8547310/) has an accessible human-cloning/clinical abstract, but its full supporting experiments were not recovered. Hypotonia and creatine deficiency do not by themselves show that GAMT performs the force-generating contractile step. The prior confident over-annotation and citation-defect interpretations are preserved historically and replaced with this explicit access limitation.
The complete cached abstracts of PMID:24415674(https://pubmed.ncbi.nlm.nih.gov/24415674/) and PMID:26319512(https://pubmed.ncbi.nlm.nih.gov/26319512/) describe functional restoration of GAMT activity by selected variants expressed in deficient human fibroblasts. The former also identifies variants with little activity. These assays corroborate catalysis; the clinical neurodevelopmental phenotype does not by itself establish that GAMT performs a distinct developmental process. Full Methods and figures were not inspected. The three ECO-linked studies, including the transcript-2 study, now have manual reference assessments and source-specific findings.
UniProt's catalytic equation explicitly identifies guanidinoacetate CHEBI:57742 and S-adenosyl-L-methionine CHEBI:59789; both structured substrates are restored to the core. The single reaction remains the basis for both core process facets.
The review YAML holds current judgments, one core and fourteen manual reference assessments. The evidence record preserves the eighteen original contexts, dates, donors and qualifiers, two sequences, four IBA placements, and exact quote locators. These original fields preserve information not all represented in the native 16-column export. It also provides a compact source manifest; repeated fetched publication headers remain available in the immutable source files and the previous full inventory. No original context or source byte is changed.
The cached MECP2 pathway was acquired because it is explicitly listed in the actual UniProt DR section. It is retained as normal source provenance, without making GAMT a transcription factor or using the pathway title as a new functional claim. Bibliography-only sources are not automatically promoted to curated references; claim-relevant evidence is assessed individually. The inherited SAM-catabolism core advisory remains deliberate, and the contraction annotation remains UNDECIDED.
The immutable earlier review contains seventeen rows and fourteen quotation associations. It lacks the partner fields needed to assign its single HuRI row uniquely to either current partner, so both possible context joins are disclosed. All original text remains in immutable Git; the public record gives its hash and precise locators instead of duplicating the full review. Both inherited reference finding statements are restored with explicit CURRENT or CORROBORATED assessments. Their short exact source anchors contain 13 words from PMID:8651275 and 4 words from UniProt, with a separate 6-word Reactome core anchor. Three new source-specific findings add exact anchors from the ECO-linked functional studies; each source remains within 25 words and no excerpt is repeated in the journal. The two finding associations remain current with disclosed excerpt adaptation, while twelve other inherited quote associations remain historical-only. The earlier journal remains an exact prefix of the appended notes, including superseded interpretations.