ProtNLM Prediction Evaluation

ProtNLM2 made no predictions (GO or subcellular location) for Cellulose synthase-like protein D3. The model returned only a protein name.

ProtNLM2 made no predictions (GO or subcellular location) for Protein kinase domain-containing protein (blue whale). The model returned only a protein name.

ProtNLM2 made no predictions (GO or subcellular location) for PrpF family protein. The model returned only a protein name.

ProtNLM2 made no predictions (GO or subcellular location) for HTH marR-type domain-containing protein. The model returned only a protein name.

ProtNLM2 made no predictions (GO or subcellular location) for Mitogen-activated protein kinase 15. The model returned only a protein name.

The ProtNLM2 API snapshot retrieved 2026-09-08 contains no GO-term predictions for Q8IRM9. RNA binding (GO:0003723) and small-subunit processome assembly (GO:0034462) are supported omission candidates: the target has the specific ABT1/Esf2 family assignment and retained RNA-recognition motif, existing PAINT/IBA annotations, and a characterized yeast Esf2 homolog that binds pre-rRNA and supports processome assembly. This is a conserved family inference, not a direct fly biochemical assay. The separate FUNCTION paragraph has CNN judgments for the broad processome/18S-rRNA claims and UNC for the exact yeast-named A0/A1/A2 cleavage-site requirements. No intrinsic nuclease activity is inferred. This completed record assesses the absence of GO output; predictions is empty because no GO or EC prediction was emitted. The narrative assessments remain in the linked function review and are not scores assigned to the omission.

The ProtNLM2 API snapshot retrieved 2026-09-08 contains no GO-term predictions for Q9VQ04. Phosphatase activity (GO:0016791) is a supported omission candidate; pseudouridine 5'-phosphatase activity (GO:1990738) is a more specific candidate supported by FlyBase's manual ortholog transfer to experimentally characterized human PUDP, together with near-complete sequence alignment and retained catalytic Asp positions. The substrate assignment is an orthology-based inference, not a direct CG5565 assay; the alignment alone does not establish substrate specificity. The separate FUNCTION claim of in-vitro D,L-glyceraldehyde 3-phosphate dephosphorylation remains UNC and is not disproved by the better-supported physiological substrate. This completed record assesses the absence of GO output; predictions is empty because no GO or EC prediction was emitted. The narrative judgment remains in the linked function review and is not a score assigned to the omission.

The ProtNLM2 API snapshot retrieved 2026-09-08 contains no GO-term predictions for Q9VB17. The retained enoyl-CoA hydratase/isomerase-like domain supports a crotonase-family assignment, but the available evidence establishes neither a diagnostic subfamily nor a characterized ortholog with matching substrate specificity. Consequently, no specific fatty-acid oxidation or catalytic GO omission is confidently identified. The existing ND root annotations record this lack of functional evidence; they are not missing positive predictions. The separate tentative FUNCTION claim about fatty-acid oxidation remains UNC, as does the ancillary mitochondrial localization claim in the narrative review. This completed record assesses the absence of GO output; predictions is empty because no GO or EC prediction was emitted. The narrative judgments remain in the linked function review and are not scores assigned to the omission.

The ProtNLM2 API snapshot retrieved 2026-09-08 contains no GO-term predictions for Q9VVS1. Mitochondrion (GO:0005739) is a supported omission candidate because the primary Dic4 study reports mitochondrial localization of the expressed Dic proteins, and the current exact accession is the native 302-residue Dic4 product. A specific transport-activity omission is not established: reconstituted Dic4 showed no detectable transport in the published dicarboxylate-carrier assays, while the accessible abstract does not establish that thiamine pyrophosphate was tested. The separate FUNCTION prediction of mitochondrial thiamine pyrophosphate uptake therefore remains UNC. This completed record assesses the absence of GO output; predictions is empty because no GO or EC prediction was emitted. The narrative judgment remains in the linked function review and is not a score assigned to the omission.

The ProtNLM2 API snapshot retrieved 2026-09-08 contains no GO-term predictions for A8Y5A1, the native 434-residue Gfat1-PF product. Omission of the complete glutamine-dependent amidotransferase activity (GO:0004360) is not a supported deficiency: this exact product lacks the catalytic cysteine and most of the glutaminase module needed for that reaction. The retained SIS domains and ligand-bound structures of homologous human isomerase domains support carbohydrate derivative binding (GO:0097367) as a broad, qualified omission candidate, not proof of target-specific binding or catalysis. Participation in UDP-N-acetylglucosamine biosynthesis remains unresolved for this isoform. The separate FUNCTION claim of the complete glutamine-dependent reaction is NPI because of the missing catalytic architecture. This completed record assesses the absence of GO output; predictions is empty because no GO or EC prediction was emitted. The narrative judgment remains in the linked function review and is not a score assigned to the omission.

ProtNLM2 emitted no GO or EC predictions for A0A0B4KEF3 in the frozen 2026-09-08 API snapshot. It emitted a protein name, a Secreted location, and the function text "Component of the rigid cuticle of the spider". The absence of GO output is an informative coverage omission: the current target is identical over all 112 residues to reviewed Lcp3 P07188 at the same FlyBase locus, which has experimental FlyBase annotations for GO:0008010 structural constituent of chitin-based larval cuticle and GO:0062129 chitin-based extracellular matrix. These are well-supported expected functions, not predictions emitted by ProtNLM. The separate function-description review supports the broad cuticle claim, leaves the rigid qualifier uncertain, and rejects the spider attribution. This record assesses missing GO output without assigning a VDCL score to an absent prediction or asserting a benchmark-wide recall value; the prediction-time input sequence and training-set membership remain unknown.

The ProtNLM2 API snapshot retrieved 2026-09-08 contains no GO-term predictions for A0A0B4LHX6. Nucleoside diphosphate kinase activity (GO:0004550) is a strongly supported omission candidate: purified Drosophila awd protein has the activity, and the selected 168-residue product contains the entire characterized 153-residue enzyme sequence, including its catalytic histidine and nucleotide-binding positions. The disagreement between FlyBase and reviewed UniProt about N-terminal initiation does not alter that conserved catalytic core; the historical prediction-time sequence is not established by the current comparison. The separate FUNCTION paragraph describing NDP-kinase phosphate transfer is CNN. This completed record assesses the absence of GO output; predictions is empty because no GO or EC prediction was emitted. The narrative judgment remains in the linked function review and is not a score assigned to the omission.

The ProtNLM2 API snapshot retrieved 2026-09-08 contains no GO-term predictions for M9NFK2. DNA-binding transcription factor activity, RNA polymerase II-specific (GO:0000981), and nucleus (GO:0005634) are supported omission candidates: primary FTZ-F1 DNA-binding/reporter experiments and betaFTZ-F1 nuclear localization complement exact sequence identity to reviewed P33244-2 and retention of the shared DNA-binding domain. This does not transfer every expression pattern of the longer isoform or establish the historical predictor input. Nuclear-receptor family membership alone does not establish ligand-regulated transcription. The separate FUNCTION paragraph supports broad transcriptional regulation but has a PLI judgment for high-affinity thyroid-hormone receptor specificity. This completed record assesses the absence of GO output; predictions is empty because no GO or EC prediction was emitted. The narrative judgments remain in the linked function review and are not scores assigned to the omission.

ProtNLM2 made no predictions (GO or subcellular location) for UDP-glycosyltransferase 73C1. The model returned only a protein name.

ProtNLM2 made no predictions (GO or subcellular location) for PGG domain-containing protein. The model returned only a protein name.

ProtNLM2 made no predictions (GO or subcellular location) for Protein kinase domain-containing protein (Xenopus). The model returned only a protein name.

ProtNLM2 made no predictions (GO or subcellular location) for ADP-ribosylation factor-like protein 1. The model returned only a protein name.