EIF2S3

Summary: The inherited tRNA-binding function is more precisely initiator methionyl-tRNA binding.

Reason: The target lies below PTN002450559 in the authenticated PTHR42854 tree, and the literal IBD carries the same contributes_to relation and yeast GCD11 experimental donor. The target MSA sequence exactly matches human P41091. Gamma forms the nucleotide/initiator-tRNA-binding centre of eIF2; donor mutagenesis and the separately assessed human/yeast disease studies support initiator-tRNA specificity. This is a refinement of a valid ancestral function, not rejection based on donor count. The imported contributes_to qualifier remains unchanged.

Proposed replacements: methionyl-initiator methionine tRNA binding

Summary: The gamma-family tRNA-binding prediction can be specified to initiator methionyl-tRNA.

Reason: The exact target carries IPR044127, the eIF2-gamma domain associated with charged-tRNA recognition. The rule is compatible with the conserved gamma role and target-specific initiation evidence. The relevant RNA is methionyl-initiator tRNA, so GO:1990856 is more informative than unrestricted tRNA binding. This domain mapping is not presented as a new human binding experiment.

Proposed replacements: methionyl-initiator methionine tRNA binding

Summary: Gamma participates in formation of the translation preinitiation complex.

Reason: The exact PTN000562153 IBD is on the human target path and is grounded in the preserved fission- and budding-yeast records. Gamma carries GTP and initiator tRNA within the eIF2 ternary complex that joins the small ribosomal subunit; it supplies part of the assembly machinery rather than merely being a downstream translation product. Human loss-of-function and complex/tRNA recovery experiments are concordant. The phylogenetic assertion is not a pairwise similarity claim.

Summary: Human patient-cell and isogenic-control experiments support cytoplasmic translation initiation.

Reason: PMID:31836389 compares EIF2S3 frameshift patient-derived iPSCs with control cells and includes CRISPR-corrected isogenic controls. Reduced polysomes and protein synthesis, altered upstream-ORF reporter translation, and impaired eIF2/initiator-tRNA complex recovery establish a translation-initiation defect. The gamma subunit performs part of the initiation machinery; selective induction of stress transcripts is compatible with reduced bulk translation. ISRIB rescues several outputs through eIF2B and is not evidence that EIF2S3 itself has nucleotide-exchange activity.

Summary: The phylogenetically supported initiation-factor activity is a principal gamma function.

Reason: The PTHR42854 tree and exact 472-residue MSA place human gamma beneath the literal PTN000562153 initiation-factor IBD. The preserved donor set includes human target experimental evidence, which is expected in PAINT and is not circular. Gamma performs nucleotide- and initiator-tRNA-dependent work in translation initiation within eIF2. This does not imply that isolated gamma recapitulates the entire heterotrimer.

Summary: Human EIF2S3 perturbation supports its initiation-factor role in HCV-IRES translation.

Reason: The original study separately targets human eIF2 gamma and eIF2B gamma. RzHCV2 and four additional gamma-directed validation ribozymes reduce HCV-IRES-driven protein output in HeLa reporter cells, with reporter RNA and cap-dependent-expression controls. The EIF2S3 result supports translation-initiation factor activity in this measured viral-IRES context. Cleavage of synthetic EIF2S3 RNA is performed by the ribozyme and is not an RNA-binding assay of the encoded gamma protein.

Summary: The eIF2-dependent initiation-factor annotation is consistent with the factor-addition experiment.

Reason: The authentic abstract reports that adding eIF2 and eIF2B improves translation in HeLa-cell extracts, including capped, uncapped and EMCV-IRES templates. The complete operative Methods and the added factor preparations were not recovered, so this paper is not described as an isolated human-gamma assay. Its curated initiation-factor claim agrees with the independently established direct role of gamma within eIF2 and can be retained with that preparation limit.

Summary: The GTPase-domain mapping matches the catalytic nucleotide cycle of eIF2 gamma.

Reason: The target has the specific eIF2-gamma GTPase architecture as well as the mapped IPR000795 domain. RHEA:19669 describes GTP hydrolysis, not guanine nucleotide exchange. Gamma contains the guanine-nucleotide site; the physiologic hydrolysis cycle is regulated by the initiation machinery, including eIF5. The shared domain is interpreted in this target context rather than as evidence for an elongation-factor function or autonomous unrestricted hydrolysis. PMID:3888990 is donor-level whole-factor corroboration, not an isolated-human-gamma assay.

Summary: The yeast gamma ortholog supports the conserved GTPase function.

Reason: The preserved supporting entity P32481 is budding-yeast GCD11. Its official experimental records include whole-eIF2 GTPase activity, while structural and mutation evidence locates the nucleotide site in gamma. The exact human/yeast family placement and conserved nucleotide-binding architecture support this ISS assertion. The donor experiment is not relabeled as purified isolated human gamma. The literal experimental donor record cites PMID:3888990; the normal cache is abstract-only and selected original text was inspected separately.

Summary: GTPase activity is coherent with gamma identity, with limits on the cited cloning study.

Reason: PMID:8106381 establishes the human 472-residue gamma cDNA and its relationship to GTP-binding translation factors; the accessible abstract also reports initiator-tRNA crosslinking. It does not provide a separately resolved human-gamma hydrolysis assay. The TAS claim is retained because gamma is the GTPase component of eIF2, corroborated by the conserved domain and donor biochemical evidence, rather than because sequence similarity alone proves an assay result.

Summary: Binding to the translation-initiation-factor subunit EIF2S1/alpha is more informative than generic protein binding.

Reason: The source has four literal positive human alpha-gamma ELISA records. The indexed original describes immobilized purified subunits and pairwise binding, while the accessible abstract specifies baculovirus-produced human eIF2. Complete operative preparation Methods were not recovered. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: Binding to the translation-initiation-factor subunit EIF2S2/beta is more informative than generic protein binding.

Reason: The source has four literal positive human beta-gamma ELISA records. Indexed original text describes pairwise binding of purified subunits; the abstract identifies baculovirus-produced human eIF2. This is stronger than inferring an interaction from co-expression, but exact preparation details remain incompletely accessible. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: Binding to the translation-initiation-factor subunit EIF2S2/beta is more informative than generic protein binding.

Reason: The literal BioPlex2 directed-edge table identifies beta bait and gamma prey. The original Methods use tagged human ORFeome baits and anti-HA affinity purification followed by mass spectrometry. This supports recovery in an associated complex, without establishing a purified binary interface. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: Binding to the translation-initiation-factor subunit EIF2S1/alpha is more informative than generic protein binding.

Reason: The original Methods separately express the three human eIF2 subunits in yeast and purify the complex by beta-His and alpha-FLAG affinity steps. Gamma recovery with alpha supports the subunit association; it is not a guanine nucleotide-exchange assay of gamma and does not isolate a binary alpha-gamma affinity. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: The reported interaction with METTL15 isoform 4 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The tested 68-279 interval belongs to the explicitly recorded alternative product A6NJ78-4. Its altered C-terminal sequence overlaps the methyltransferase region; the saved evidence does not establish a specific active-enzyme class for this assayed product. The positive interaction remains valid without an assertion that the fragment is inactive. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. PER1 is a clock transcriptional coregulator, but no independent mechanism linking this EIF2S3 pair to transcriptional coregulation is established. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The reported interaction with SYNCRIP isoform 4 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. SYNCRIP is an RNA-binding protein, and the recorded alternative product retains substantial RNA-recognition architecture. That identity alone does not establish a more informative supported binding MF for this pair. No RNA-binding activity is transferred to gamma. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. DXO is an RNA decapping/exonuclease enzyme, but no independent mechanistic or established-complex evidence links this screen pair to that enzymatic role. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The reported interaction with CDKN1A is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. CDKN1A is a cyclin-dependent kinase inhibitor, not a protein kinase. The tested interaction does not identify a supported more specific binding class; kinase binding would misclassify the partner. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with HTT residues 1-546 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The source explicitly tests an N-terminal huntingtin fragment. The positive reporter interaction does not establish the functions of intact huntingtin in this fragment or identify a justified specific binding MF. Native complex formation remains unresolved. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. UBE2A is a ubiquitin-conjugating E2 enzyme. The screen does not establish an EIF2S3-UBE2A ubiquitination mechanism or an independently established functional complex. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. GABPB1 supplies transcriptional activation in the GABP complex, and the recorded isoform-2 change preserves the mapped activation region. This class information does not establish a functional transcriptional interaction with EIF2S3. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. GABPB1 is the activation component of GABP; the isoform-3 alteration is recorded explicitly. No independent mechanism or established complex connects this screen pair to that coregulator activity. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The reported interaction with INCA1 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. INCA1 is described as a cyclin-dependent kinase inhibitor. The positive pair does not establish a kinase partner or an informative supported alternative binding MF; inhibition of a kinase does not make INCA1 a kinase. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with ZNF829 isoform 3 residues 82-513 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The literal record identifies an alternative-product fragment. The saved official record gives a predicted transcriptional role, but the bounded evidence does not establish the assayed product as a sequence-specific DNA-binding transcription factor. Generic binding is retained without inventing that class. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with SPATS1 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The exact positive pair is preserved. The captured official evidence does not establish a suitable specific partner class, and a testis-associated name alone does not supply one. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with FASTKD1 residues 173-847 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. FASTKD1 is an RNA-associated mitochondrial regulatory protein. Its name does not establish protein-kinase activity. The fragment-specific positive assay is retained without an unsupported kinase-binding refinement or an assertion of a native cytosolic complex. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with MAPDA isoform 2 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The recorded Q6DHV7-2 product has a major C-terminal replacement within the canonical deaminase architecture. The active-site position retained in the reference sequence is insufficient by itself to resolve the assayed alternative product's specific enzyme class. This class uncertainty does not negate the positive pair or demonstrate inactivity. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. The assayed OTUD7B interval retains its deubiquitinase domain and catalytic residues. That credible enzyme class does not itself establish a functional deubiquitination-related interaction with EIF2S3. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The reported interaction with SPATA12 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The positive source-specific reporter interaction is retained. The saved functional evidence does not establish an informative supported binding class; no reproductive biological process is inferred from the partner name. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. TAF8 is a TFIID component, and its recorded alternative-product change does not remove the mapped histone-fold domain. EIF2S3 membership in TFIID or an independent transcription-initiation mechanism is not established by this pair. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The reported interaction with PELI3 isoform 4 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The source names Q8N2H9-4, whose replacement and truncation remove the canonical C-terminal RING architecture. The E3 class of other PELI3 products cannot simply be assigned to this tested product. Retain the positive interaction with unresolved informative class, without asserting that the fragment has no biological activity. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. BAHD1 has independently supported transcriptional corepressor activity, but the screen does not establish that its EIF2S3 interaction operates in that mechanism or an independently verified complex. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The reported interaction with LARP4B is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. LARP4B is an RNA-binding translational regulator, but the bounded source evidence does not establish a specific initiation-factor or other supported binding class for this pair. Its RNA activity is not an activity of gamma merely because they associate. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with DPF1 isoform 2 residues 28-414 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The tested fragment is retained exactly. DPF1 is a chromatin-remodelling-complex component; the captured similarity-based cofactor description does not establish this fragment as a sequence-specific DNA-binding transcription factor. No unsupported TF-binding refinement is made. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with ASIC4 isoform 4 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The official record distinguishes ASIC4 from channels with demonstrated autonomous acid-gated conductance. The captured alternative-product interaction does not resolve transporter activity or a particular transport-complex role. Generic binding is retained as a positive assay result, not rejected because of membrane topology. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. MAP3K5 has independently established protein-kinase activity. The source does not establish a kinase-substrate mechanism or a functional kinase complex for this particular EIF2S3 pair; the separate retracted AKT-regulation report is not used as evidence. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. ZKSCAN3 has independently supported sequence-specific DNA-binding transcription-factor activity. The screen does not establish a functional transcriptional association with EIF2S3 or an endogenous nuclear complex. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The verified screen interaction is retained as a non-core binding observation.

Reason: PMID:32814053 reports the exact pair in a yeast two-hybrid matrix with reporter retesting; its literal alternative-product identifiers and fragment intervals remain unchanged. Array, pooling and validated records are stages of that workflow rather than independent assay classes. POLR3F is an RNA polymerase III component. That complex identity does not independently place EIF2S3 in the polymerase machinery or demonstrate a functional polymerase-binding mechanism. Under the campaign rule, the positive interaction remains KEEP_AS_NON_CORE; partner class alone is insufficient for a functional refinement. This is not a false-hit judgment, a compartment-impossibility claim or a requirement for a purified binary assay.

Summary: The reported interaction with RANGRF isoform 3 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The exact alternative product is preserved. RANGRF regulates RAN nucleotide handling, but nucleotide release or inhibition of loading does not automatically establish guanine-nucleotide exchange factor activity. No unsupported GEF-binding class is inferred. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with RASSF1 isoform 4 is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The record is specific to an alternative product with potentially different domain and interaction properties. The bounded evidence does not establish a particular scaffold or enzyme class for this assayed product; the positive pair is retained. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: The reported interaction with WBP1L is retained as a non-core binding observation.

Reason: PMID:32814053 uses a yeast two-hybrid matrix with reporter retesting; the exact pair, alternative-product identifier and any fragment interval are preserved in the literal IntAct records. Array, pooling and validated records are stages of this workflow, not three independent assay classes. The literal positive interaction is retained. The captured evidence supplies no informative supported partner-class MF, and the WW-domain-binding name does not identify the interface tested here. This does not establish an endogenous human complex or an orthogonal mammalian interaction assay for this pair.

Summary: CDC123 interaction supports generic non-core binding; the narrower chaperone-binding assignment remains unestablished.

Reason: The literal BioPlex3 IntAct record reports affinity-capture association with CDC123. No matching edge was recovered in the selected 293T table; that absence is neither a negative result nor proof of an HCT116-specific assay. Current UniProt classifies CDC123 as an ATP-dependent protein-folding chaperone, and the human interface/assembly evidence is positive. However, these inspected assays do not establish the nascent or unfolded client-folding or transport criterion of GO:0051087. A folded domain structure alone does not disprove such a role. Retain supported generic non-core binding without promoting the assembly model to that narrower molecular function.

Summary: Binding to the translation-initiation-factor subunit EIF2S1/alpha is more informative than generic protein binding.

Reason: Two literal IntAct records and the BioPlex3 293T table support alpha-gamma affinity-capture association. The original uses tagged human baits and mass spectrometry, with 293T and HCT116 datasets kept distinct. The selected alpha-bait edge is present in the 293T table. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: Binding to the translation-initiation-factor subunit EIF2S2/beta is more informative than generic protein binding.

Reason: Two literal IntAct records and the BioPlex3 293T table support beta-gamma affinity-capture association. The selected beta-bait edge is a positive 293T result; the broader study also profiles HCT116 and does not make these cell contexts interchangeable. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: CDC123 interaction supports generic non-core binding; the narrower chaperone-binding assignment remains unestablished.

Reason: The OpenCell gamma-bait record recovers H7BZW7;O75794;X6RF82 in human HEK293T cells. Endogenous tagging and nanobody affinity capture support association while preserving the grouped identification. Current UniProt classifies CDC123 as an ATP-dependent protein-folding chaperone, and the human interface/assembly evidence is positive. However, these inspected assays do not establish the nascent or unfolded client-folding or transport criterion of GO:0051087. A folded domain structure alone does not disprove such a role. Retain supported generic non-core binding without promoting the assembly model to that narrower molecular function.

Summary: Binding to the translation-initiation-factor subunit EIF2S1/alpha is more informative than generic protein binding.

Reason: The saved OpenCell CSV contains the gamma-bait alpha-prey edge. Endogenous split-fluorescent tagging and nanobody affinity capture in HEK293T cells recover associated proteins. The literal alpha group includes P05198 with two additional accessions; the grouped mass-spectrometry identification is retained without resolving an unreported natural isoform. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: Binding to the translation-initiation-factor subunit EIF2S2/beta is more informative than generic protein binding.

Reason: The saved OpenCell CSV contains a gamma-bait beta-prey P20042 edge. Endogenous tagging and nanobody affinity capture support an association in human HEK293T cells. Recovery of the eIF2 subunit complex does not by itself establish an isolated binary dissociation constant. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: CDC123 interaction supports generic non-core binding; the narrower chaperone-binding assignment remains unestablished.

Reason: The source resolves a physical human CDC123 interface with the engineered human gamma 362-471 fragment and tests assembly-related human constructs in yeast. The fragment is not a natural product; assembly-order interpretation retains its organism and assay limits. Current UniProt classifies CDC123 as an ATP-dependent protein-folding chaperone, and the human interface/assembly evidence is positive. However, these inspected assays do not establish the nascent or unfolded client-folding or transport criterion of GO:0051087. A folded domain structure alone does not disprove such a role. Retain supported generic non-core binding without promoting the assembly model to that narrower molecular function.

Summary: Binding to the translation-initiation-factor subunit EIF2S1/alpha is more informative than generic protein binding.

Reason: The original human alpha-gamma yeast two-hybrid assay has a weak baseline signal that is enhanced by human CDC123. This is distinct from Figure 4B, which uses human CDC123 with yeast eIF2 subunits. Retain the measured human pair and the heterologous reporter context without converting either assay into purified human binary affinity. Alpha and beta are established components of the translation initiation factor eIF2, so GO:0031369 describes the partner class. This refinement does not transfer the partner's activity to gamma.

Proposed replacements: translation initiation factor binding

Summary: CDC123 interaction supports generic non-core binding; the narrower chaperone-binding assignment remains unestablished.

Reason: The physical AP-MS network contains a CDC123-bait to EIF2S3-prey edge from tagged human ORFeome baits in U2OS cells and anti-HA capture. This is association evidence distinct from the integrated computational model. Current UniProt classifies CDC123 as an ATP-dependent protein-folding chaperone, and the human interface/assembly evidence is positive. However, these inspected assays do not establish the nascent or unfolded client-folding or transport criterion of GO:0051087. A folded domain structure alone does not disprove such a role. Retain supported generic non-core binding without promoting the assembly model to that narrower molecular function.

Summary: GTP binding is supported by the gamma nucleotide-binding architecture.

Reason: The literal InterPro inputs include the GTP-binding translation-factor domain and the specific eIF2-gamma domain. Together with the target sequence and conserved nucleotide cycle, they support GTP binding. The EF-Tu-like shared fold does not imply that human EIF2S3 is a translation elongation factor, and GTP binding is distinct from exchange catalysis.

Summary: The negated nuclear annotation is supported within the reported steady-state localization assay.

Reason: The imported assertion is NOT nucleus. PMID:12426392 expresses human eIF2 gamma-GFP in BHK cells and reports strong nuclear depletion, with a nuclear-to-cytoplasmic concentration ratio of approximately 0.013. The source also tests leptomycin B and an ectopic nuclear-localization sequence; gamma differs from beta in its response. Retain the curated negative with this cellular and construct scope rather than interpreting it as proof that no gamma molecule can ever enter a nucleus.

Summary: The positive cytoplasmic localization can be refined to cytosol.

Reason: PMID:12426392 directly resolves human gamma-GFP enrichment in the cytoplasm of BHK cells and strong nuclear depletion; that imaging experiment alone is not described as cytosol fractionation. Independent target-resolved eIF2-complex and curated Reactome component evidence places human P41091 in the cytosol. Under the correct-but-general refinement rule, cytosol is the more informative supported location while preserving the original assay and source fields.

Proposed replacements: cytosol

Summary: The UniProt location mapping agrees with the cytosolic eIF2 pool.

Reason: The literal SL-0091 mapping is cytosol. Human gamma participates in cytosolic eIF2 and initiation complexes, independently supported by source-resolved biochemical associations and curated component paths. This is not inferred merely from the absence of a secretion or membrane signal.

Summary: The cytosolic gamma component is explicitly represented during circularized ceruloplasmin-mRNA initiation-complex formation.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-156808 -> R-HSA-72571 -> R-HSA-72532 -> R-HSA-72531 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The component-location assertion is retained without assigning every activity in the reaction to gamma.

Summary: The cytosolic gamma component is explicitly represented during the L13a-associated ceruloplasmin-mRNA complex.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-156823 -> R-HSA-156804 -> R-HSA-72571 -> R-HSA-72532 -> R-HSA-72531 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The localization does not by itself assign L13a-mediated translational repression to gamma.

Summary: The cytosolic gamma component is explicitly represented during initiation on non-circularized mRNA.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-157849 -> R-HSA-72571 -> R-HSA-72532 -> R-HSA-72531 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The component-location assertion is retained without assigning every activity in the reaction to gamma.

Summary: The cytosolic gamma component is explicitly represented during PERK phosphorylation of the eIF2 alpha subunit.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-381111 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The ER-membrane compartment belongs to the PERK kinase; the eIF2 target complex is cytosolic.

Summary: The cytosolic gamma component is explicitly represented during ABCF1 association with eIF2.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-5227009 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The pathway heading does not establish ABC-transporter activity in gamma.

Summary: The cytosolic gamma component is explicitly represented during GTP hydrolysis and release of initiation factors.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72619 -> R-HSA-72530 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The event model retains the distinct functions of eIF2 and the eIF5 GTPase-activating machinery.

Summary: The cytosolic gamma component is explicitly represented during ribosomal scanning.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72621 -> R-HSA-72592 -> R-HSA-72571 -> R-HSA-72532 -> R-HSA-72531 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The scanning reaction's eIF1 catalyst is distinct from the gamma-containing complex.

Summary: The cytosolic gamma component is explicitly represented during formation of eIF2:GTP.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72663 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The component-location assertion is retained without assigning every activity in the reaction to gamma.

Summary: The cytosolic gamma component is explicitly represented during initiator-tRNA ternary-complex formation.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72669 -> R-HSA-72531 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The component-location assertion is retained without assigning every activity in the reaction to gamma.

Summary: The cytosolic gamma component is explicitly represented during the GDP-eIF2/eIF2B intermediate.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72670 -> R-HSA-72530 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The component-location assertion is retained without assigning every activity in the reaction to gamma.

Summary: The cytosolic gamma component is explicitly represented during 43S preinitiation-complex formation.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72691 -> R-HSA-72532 -> R-HSA-72531 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The component-location assertion is retained without assigning every activity in the reaction to gamma.

Summary: The cytosolic gamma component is explicitly represented during start-codon recognition.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72697 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The component-location assertion is retained without assigning every activity in the reaction to gamma.

Summary: The cytosolic gamma component is explicitly represented during eIF2 activation by eIF2B.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-72722 -> R-HSA-72529 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The exchange catalyst is eIF2B, not EIF2S3.

Summary: The cytosolic gamma component is explicitly represented during PKR phosphorylation of the eIF2 alpha subunit.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-9835885 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. PKR supplies the kinase activity, and alpha is the phosphorylated eIF2 subunit.

Summary: The cytosolic gamma component is explicitly represented during DELE1-HRI-enhanced phosphorylation of the eIF2 alpha subunit.

Reason: The saved official Reactome pages provide the literal input/output and component path R-HSA-9840415 -> R-HSA-72515 -> R-HSA-72514. The terminal human entity R-HSA-72514 maps to UniProt P41091, with the processed chain 2-472, in cytosol. This verifies the target's location within the curated complex rather than inferring it from an event title. The mitochondrial-stress origin of the signal does not place gamma inside mitochondria.

Summary: The ancestral eIF2-complex membership is supported by the target phylogeny and human experiments.

Reason: The literal PTN000562154 IBD lies on the target path in the 345-node tree, whose target MSA sequence exactly matches human P41091. Yeast and human experimental complex evidence underlies the placement. Human gamma occurring among the descendants that ground the IBD is expected, not circular propagation. Direct human alpha/beta association independently supports eIF2-complex membership.

Summary: Human gamma associates with alpha and beta as a constituent of eIF2.

Reason: PMID:23063529 directly expresses tagged wild-type and I222T human eIF2 gamma in HeLa cells. Endogenous alpha and beta co-precipitate with wild-type gamma; the mutant selectively impairs beta recovery. This directly supports human gamma membership in the heterotrimer. The parallel yeast mutations and archaeal structural comparison are distinct evidence and are not presented as human assays.

Summary: The source-specific interaction evidence supports eIF2-complex membership.

Reason: The human gamma affinity-capture experiment in PMID:23063529 recovers the endogenous alpha and beta subunits with tagged wild-type gamma, while I222T reduces beta association. This supports part_of the eukaryotic translation initiation factor 2 complex. The human co-precipitation result is distinct from the yeast growth and start-codon-fidelity assays.

Summary: Human-cell co-immunoprecipitation supports the gamma-alpha association in eIF2.

Reason: PMID:31836389 measures endogenous eIF2 subunit co-immunoprecipitation in control and EIF2S3-frameshift patient iPSCs. The mutation impairs alpha incorporation, and alpha overexpression improves complex and initiator-tRNA recovery. The positive wild-type and rescue observations support eIF2 membership; they do not establish a purified binary gamma-alpha affinity or assign CDC123 assembly activity to gamma.

Summary: The supported initiation process is specifically cytoplasmic.

Reason: The human HeLa HCV-IRES reporter experiments in PMID:10900014 support EIF2S3 participation in translation initiation. The eIF2-dependent initiation machinery and measured reporter output concern cytoplasmic ribosomes, so GO:0002183 is the more informative process term. This refinement preserves the viral-IRES assay context and does not infer viral entry, RNA cleavage by gamma, or nuclear translation.

Proposed replacements: cytoplasmic translational initiation

Summary: The gamma-specific regulatory claim remains unresolved from the accessible source.

Reason: The authentic PMID:18639529 abstract reports bacterially expressed human eIF2 subunits, pairwise associations and interactions with regulators; its mechanistic emphasis includes beta binding to eIF2B, kinases and other regulators. The operative full Methods and Results were not recovered. The general role of eIF2 in regulated initiation is positive, but the available material does not resolve which gamma activity specifically supports this separate regulation annotation. Retain the curator assertion as uncertain rather than infer misattribution from the abstract or transfer beta-specific mechanisms.

Summary: The source supports translation initiation rather than a demonstrated RNA-contact activity.

Reason: PMID:10900014 tests human EIF2S3-dependent HCV-IRES translation in cells. Its ribozymes cleave EIF2S3 RNA; the study does not measure gamma binding to the IRES or another RNA. GO:0003743 captures the supported initiation-factor role. This is an explicitly lateral biological correction, not an ontology descendant refinement: current QuickGO is_a/part_of ancestry does not place GO:0003743 under GO:0008135. Independent tRNA-binding evidence remains relevant to gamma elsewhere, but is not relabeled as the assay performed in this source. The existing initiation-factor annotation from this paper is retained as a distinct imported context, with no new row proposed.

Proposed replacements: translation initiation factor activity

Summary: The reported initiator-tRNA contact supports an explicit lateral correction to methionyl-initiator methionine tRNA binding.

Reason: The authentic PMID:8106381 abstract reports beta/gamma crosslinking to Met-tRNAi with two reagents, alongside human gamma cloning and expression. Complete preparation and species details for the crosslinking assay were not recovered. The observed RNA-contact role supports GO:1990856 without assigning direct mRNA- or IRES-binding activity or an isolated-human-subunit affinity measurement. This replacement is lateral, not a descendant refinement of GO:0008135 in the current ontology. It differs from the PMID:10900014 perturbation-only context because this source reports a physical tRNA-contact experiment.

Proposed replacements: methionyl-initiator methionine tRNA binding

Summary: The cadherin-binding assertion is not resolved by the available proximity-proteomics evidence.

Reason: PMID:25468996 uses E-cadherin-BirA proximity labeling in human MKN28 cells and distinguishes a broad proximity proteome from a smaller set of junctional-localization validations. The main text and Methods establish that design, but the target-specific supplementary table and any EIF2S3-specific orthogonal cadherin-binding result were not recovered. Proximity can identify a nearby translation factor without demonstrating cadherin contact. Preserve the imported HDA assertion as unresolved rather than declare nonbinding or silently convert a proximity assay into binary binding.

Summary: The source-specific extracellular-exosome detection remains unverified.

Reason: The authentic PMID:20458337 abstract reports a human B-cell exosome proteome and a smaller MHC-II-associated subset. The exact EIF2S3 identification, peptide group and relevant supplementary table were not recovered; attempted supplementary downloads are retained as failures. Cytosolic function does not exclude extracellular-vesicle cargo, so no biological absence or contamination verdict follows. The curator's positive source context is retained as uncertain pending the target-level evidence.

Summary: The conserved gamma role supports binding to methionyl-initiator tRNA.

Reason: The literal ISS donor is budding-yeast GCD11/P32481. Its current experimental records and the authenticated target phylogeny agree with gamma contributing the initiator-tRNA-binding site. The independently assessed disease study measures altered initiator-tRNA binding in purified yeast eIF2 and separately examines human-cell translation; these assay contexts are not merged into an isolated human-gamma affinity measurement. The specific RNA-binding function is coherent across these sources.