BioReason-Pro SFT Prediction Evaluation

GO:0003824 (catalytic activity) is a very generic ancestor of GO:0004146 (dihydrofolate reductase activity), which is the specific molecular function of dfrP in GOA. LSP because a more specific term is already annotated. The prediction is technically correct but uninformative.

CNN. GO:0016491 (oxidoreductase activity) is in GOA for dfrP via IEA. The curated review notes this should be REMOVED as redundant with the more specific GO:0004146 (dihydrofolate reductase activity), which is already annotated. CNN because the term is in training data, though LSP relative to DHFR activity.

GO:0016645 is an oxidoreductase hierarchy term (acting on CH-NH group of donors) that is less specific than GO:0004146 (dihydrofolate reductase activity). DHFR reduces dihydrofolate (acting on the CH-NH group) using NADPH. LSP because GO:0004146 is the specific and annotated term in GOA.

DHFR uses NADPH as cofactor (NADP as acceptor during reverse reaction). GO:0050661 (NADP binding) is in GOA for dfrP via IEA. GO:0016646 (oxidoreductase with NAD or NADP as acceptor) is a related hierarchy term. CNN because NADP binding is in training data, though GO:0016646 is a less specific hierarchy term than GO:0004146.

ACCEPT in curated review. GO:0004146 is the core molecular function of dfrP - the phage-encoded DHFR catalyzes NADPH-dependent reduction of dihydrofolate to tetrahydrofolate. GO:0004146 is in GOA for dfrP via IEA (InterPro IPR001796, EC 1.5.1.3). CNN because the term is already in training data.

GO:0050896 (response to stimulus) is an extremely generic term at the top of the GO biological process hierarchy. dfrP does not have a specific response function; it is an enzyme. LSP because this prediction adds no useful information and is several levels above any relevant process term.

GO:0042221 (response to chemical) is a generic term. dfrP encodes a trimethoprim-resistant DHFR that can confer antibiotic resistance to the infected host. While resistance to trimethoprim (a chemical inhibitor of DHFR) is biologically relevant, GO:0042221 is too broad and GO:0009410 (response to xenobiotic stimulus) or more specific terms would better capture the antibiotic resistance function. LSP.

dfrP encodes a trimethoprim-resistant DHFR of the DfrA family that can confer antibiotic resistance. Response to xenobiotic stimulus (trimethoprim being a xenobiotic antibiotic) is a biologically relevant annotation. While not explicitly in GOA for dfrP, the trimethoprim-resistance function is in the training data context. CNN based on the antibiotic resistance function documented in the literature.

dfrP encodes a soluble DHFR enzyme with no transmembrane domains or signal peptides. Cytoplasmic localization is consistent with its enzymatic function in folate metabolism. While cytoplasm is not explicitly in GOA for dfrP, the enzyme's soluble nature and cytoplasmic substrate (dihydrofolate) are documented. CNN because the cytoplasmic localization of DHFR enzymes is in the training data context for this well-characterized enzyme family.

ACCEPT in curated review. Fibrolase is a zinc metalloendopeptidase (EC 3.4.24.72) that cleaves internal peptide bonds in fibrin and fibrinogen at X-Leu bonds. GO:0004222 is in GOA for fibrolase via IEA (InterPro IPR001590, EC 3.4.24.72). CNN because the term is already in training data.

Fibrolase is a zinc metalloproteinase (1 mol Zn/mol protein, PMID:1898066), not a calcium-binding protein. GO:0005509 (calcium ion binding) is not in GOA for fibrolase. The metal ion binding annotation in GOA is GO:0046872 (metal ion binding) referring to the catalytic zinc. Calcium binding is not documented for fibrolase and is biologically inconsistent with its mechanism as a zinc-dependent endopeptidase. This prediction confuses fibrolase with calcium-dependent enzymes.

Fibrolase is a zinc metalloproteinase that requires zinc (not magnesium) for catalytic activity. GO:0000287 (magnesium ion binding) is not in GOA for fibrolase. The relevant metal is zinc, annotated via GO:0046872 (metal ion binding). Magnesium binding is not documented for fibrolase. NPI because magnesium is the wrong metal for this zinc-dependent metalloendopeptidase.

ACCEPT in curated review. Fibrolase is a fibrinolytic metalloproteinase that directly hydrolyzes fibrin and fibrinogen. GO:0006508 is in GOA for fibrolase via IEA (InterPro IPR001590). CNN because proteolysis is in training data.

ACCEPT in curated review context. Fibrolase directly cleaves fibrin and fibrinogen, disrupting blood coagulation/clot formation in envenomated prey. GO:0035821 (modulation of process of another organism) is in GOA for fibrolase via IEA. The fibrinolytic activity constitutes regulation (specifically negative regulation) of blood coagulation in the prey organism. CNN because the fibrinolytic/ anticoagulant function is established in training data, though GO:0035821 is the more specific annotated term.

ACCEPT in curated review. Fibrolase is a secreted venom protein that acts extracellularly on fibrin/fibrinogen in the victim's blood. GO:0005576 is in GOA for fibrolase via IEA (UniProt subcellular location SL-0243). CNN because extracellular localization is already in training data.

Protein-containing complex binding; DNAAF1 binds dynein complexes during assembly. In GOA.

Positive regulation of gene expression; DNAAF1 is a dynein assembly factor, not a gene expression regulator. The curated review accepts dynein complex binding and axoneme assembly, NOT gene expression regulation. This is an incorrect prediction likely from confusion with different DNAAF1 literature.

Negative regulation of smoothened signaling; DNAAF1 is a dynein arm assembly factor with no documented role in Hedgehog/smoothened signaling. Incorrect prediction.

Regulation of calcium-mediated signaling; DNAAF1 is a dynein assembly factor. No documented role in calcium signaling. Incorrect prediction.

Axo-dendritic transport; cilia/flagella with properly assembled dynein arms may contribute to neuronal transport, but this is not a documented primary DNAAF1 function. UNC.

Negative regulation of ERK1/ERK2 cascade; DNAAF1 is a dynein assembly factor with no documented role in MAPK signaling. Incorrect prediction from training data contamination with unrelated pathways.

Cytoplasm; DNAAF1 functions in the cytoplasm where dynein pre-assembly occurs. NEW in curated review (correctly predicted). In GOA.

Nucleosome binding; ACCEPT in curated review. NPAC/GLYR1 contains a PWWP domain that binds nucleosomes. In GOA.

DNA binding; ACCEPT in curated review. NPAC binds chromatin/DNA. In GOA.

Positive regulation of histone acetylation; ACCEPT in curated review as GO:0140673. NPAC promotes histone acetylation during transcription elongation. In GOA.

Positive regulation of transcription by RNA Pol II; ACCEPT in curated review. NPAC facilitates Pol II transcription elongation. In GOA.

Nucleus; ACCEPT in curated review. NPAC is a nuclear protein. In GOA.

Catalytic activity; too general for PGRP-LB. More specific: N-acetylmuramoyl-L-alanine amidase activity.

Peptidoglycan muralytic activity; ACCEPT in curated review. PGRP-LB cleaves peptidoglycan. In GOA.

Hydrolase activity; too general.

Protein binding; present in GOA. PGRP-LB interacts with IRAK/IMD pathway components.

N-acetylmuramoyl-L-alanine amidase activity; ACCEPT in curated review. Core PGRP-LB function. In GOA.

Acting on carbon-nitrogen bonds; amidase mechanism. In GOA.

In linear amides; amidase mechanism. In GOA.

Biological process in interspecies interaction; PGRP-LB functions in host-pathogen interaction. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Catabolic process; peptidoglycan catabolism. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Response to biotic stimulus; bacteria. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Defense response; ACCEPT in curated review. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Negative regulation of peptidoglycan recognition protein signaling; NEW/ACCEPT in curated review. Core PGRP-LB function: amidase scavenges PGN to prevent excess immune activation. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Response to bacterium; PGRP-LB responds to bacterial PGN. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Aminoglycan metabolic process; peptidoglycan is an aminoglycan. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Defense response to bacterium; in curated review. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Aminoglycan catabolic process; PGRP-LB degrades PGN. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Glycosaminoglycan metabolic process; peptidoglycan is NOT a glycosaminoglycan. Peptidoglycan is a distinct polymer (N-acetylmuramic acid + N-acetylglucosamine + peptides). This is an incorrect prediction confusing peptidoglycan with GAGs.

Defense response to Gram-positive bacterium; MODIFY in curated review. In GOA.

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

High-level GO hierarchy term; less specific than core PGRP-LB functions (N-acetylmuramoyl-L-alanine amidase, negative regulation of immune signaling via peptidoglycan degradation).

Peptidoglycan metabolic process; ACCEPT in curated review. In GOA.

Glycosaminoglycan catabolic process; peptidoglycan is NOT a glycosaminoglycan. PGRP-LB degrades peptidoglycan, not GAGs. Incorrect prediction.

Extracellular region; MODIFY in curated review (PGRP-LB is extracellular/secreted). In GOA.

Heme binding; ACCEPT in curated review. TDO is a heme-dependent dioxygenase. In GOA.

Tryptophan 2,3-dioxygenase activity; ACCEPT in curated review. Core TDO function. In GOA.

Tryptophan catabolic process to kynurenine; ACCEPT (as GO:0006569 L-tryptophan catabolic process + GO:0006727 ommochrome) in curated review. Core TDO pathway. In GOA.

Catalytic complex; TDO forms a homotetrameric complex. In GOA.

Cytosol; ACCEPT in curated review. TDO is a cytosolic enzyme. In GOA.

Metal ion binding; ACCEPT in curated review. CCoAOMT enzymes require divalent metal ions. In GOA.

Caffeoyl-CoA O-methyltransferase activity; core predicted function. ACCEPT in curated review. In GOA.

Phenylpropanoid biosynthetic process; ACCEPT in curated review. CCoAOMT functions in the phenylpropanoid pathway. In GOA.

Cytosol; ACCEPT in curated review. CCoAOMT enzymes are cytosolic. In GOA.

Too general for BRI1; less specific than the accepted core functions (brassinosteroid receptor kinase activity).

Too general for BRI1; less specific than the accepted core functions (brassinosteroid receptor kinase activity).

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Too general for BRI1; less specific than the accepted core functions (brassinosteroid receptor kinase activity).

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Correct prediction for BRI1, consistent with curated review annotations. Already in GOA training data for this well-characterized brassinosteroid receptor.

Protein binding is too generic and not a specific CHL1 function. CHL1 is a TN immune receptor (TIR-NBS), not annotated for protein binding. The curated review does not include protein binding. FREQUENCY_BIAS.

Positive regulation of cell death; CHL1 is involved in defense and potentially HR (hypersensitive response) which involves cell death, but the curated review does not annotate this directly. UNC.

Defense response to fungus; CHL1 is an immune receptor involved in defense. ACCEPT in curated review. In GOA.

Defense response to bacterium; CHL1 mediates defense against bacterial pathogens. ACCEPT in curated review. In GOA.

Mitochondrion; CHL1 is a TN (TIR-NBS) immune receptor in the cytoplasm. The curated review annotates cytoplasm (ACCEPT) not mitochondrion. This is an incorrect localization prediction. CHL1 does NOT localize to mitochondria.

Identical protein binding; COP1 homodimerizes. ACCEPT in curated review. In GOA.

Ubiquitin protein ligase activity; ACCEPT in curated review. Core COP1 E3 function. In GOA.

Photoperiodism, flowering; KEEP_AS_NON_CORE in curated review. COP1 regulates flowering time. In GOA.

Anthocyanin-containing compound metabolic process; KEEP_AS_NON_CORE in curated review. COP1 represses anthocyanin biosynthesis genes. In GOA.

Photomorphogenesis; ACCEPT in curated review. Core COP1 function (represses photomorphogenesis in dark). In GOA.

Regulation of stomatal movement; KEEP_AS_NON_CORE in curated review. COP1 regulates stomatal development. In GOA.

Entrainment of circadian clock; KEEP_AS_NON_CORE in curated review. COP1 has circadian roles. In GOA.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0060928 is 'atrioventricular node cell development'. The canonical GO term is taxonomically incompatible with this gene product. The pair is classified NPI; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: The raw label describes flavonoid biosynthesis, but GO:0060928 is atrioventricular node cell development. That animal cardiac-development term is taxonomically impossible for Arabidopsis COP1, so the malformed pair is NPI when assessed by its GO ID. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0060928 resolves to 'atrioventricular node cell development', not 'positive regulation of flavonoid biosynthetic process'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Current-snapshot audit reclassified CNN to NPI because all exact AIGR actions are negative (REMOVE). This exact-join decision supersedes conflicting conclusions in the earlier rationale. Retained biological rationale. DNA repair; UNDECIDED in curated review. COP1 has been implicated in DNA damage response. In GOA.

Response to UV-B; ACCEPT in curated review. COP1 mediates UV-B responses. In GOA.

Response to absence of light; ACCEPT in curated review (skotomorphogenesis). Core COP1 function. In GOA.

Nuclear body; ACCEPT in curated review. COP1 localizes to nuclear bodies (photobodies). In GOA.

Nuclear ubiquitin ligase complex; ACCEPT in curated review. COP1 is part of COP1-SPA nuclear E3 complex. In GOA.

Ethylene binding; ACCEPT in curated review. ETR1 N-terminal transmembrane domain binds ethylene with a copper cofactor. Core MF. In GOA.

Identical protein binding; ETR1 homodimerizes. ACCEPT in curated review. In GOA.

Protein histidine kinase activity; ACCEPT in curated review. ETR1 has two-component kinase domain. In GOA.

Seed dormancy process; KEEP_AS_NON_CORE in curated review. ETR1 regulates seed germination via ethylene. In GOA.

Cell division; KEEP_AS_NON_CORE in curated review. Ethylene regulates cell division in some contexts. In GOA.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0009731 is 'detection of sucrose stimulus'. The canonical GO ID denotes a different, unsupported function from the raw label. The pair is classified NPI; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: The raw label describes ethylene detection, but the supplied GO:0009731 identifier is detection of sucrose stimulus. ETR1 is an ethylene receptor, not a sucrose sensor. The malformed pair is NPI when assessed by its GO ID; the raw label remains preserved for provenance. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0009731 resolves to 'detection of sucrose stimulus', not 'detection of ethylene'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Cytokinin metabolic process; KEEP_AS_NON_CORE in curated review. ETR1 interacts with cytokinin signaling. In GOA.

Response to heat; KEEP_AS_NON_CORE in curated review. ETR1 is involved in heat stress responses. In GOA.

Regulation of stomatal movement; KEEP_AS_NON_CORE in curated review. Ethylene regulates stomata. In GOA.

Response to insect; KEEP_AS_NON_CORE in curated review. Ethylene signaling in defense. In GOA.

Sugar mediated signaling pathway; KEEP_AS_NON_CORE in curated review. In GOA.

Phloem or xylem histogenesis; KEEP_AS_NON_CORE in curated review. Ethylene regulates vascular development. In GOA.

Response to molecule of bacterial origin; KEEP_AS_NON_CORE in curated review. Pathogen-produced ethylene activates ETR1. In GOA.

Response to salt stress; KEEP_AS_NON_CORE in curated review. Ethylene signaling in salt stress. In GOA.

Regulation of seedling development; KEEP_AS_NON_CORE in curated review. In GOA.

Defense response to bacterium; KEEP_AS_NON_CORE in curated review. Ethylene in defense. In GOA.

Current-snapshot audit reclassified CNN to NPI because all exact AIGR actions are negative (MARK_AS_OVER_ANNOTATED). This exact-join decision supersedes conflicting conclusions in the earlier rationale. Retained biological rationale. Hydrogen peroxide biosynthetic process; MARK_AS_OVER_ANNOTATED in curated review. In GOA.

Response to auxin; KEEP_AS_NON_CORE in curated review. Cross-talk between ethylene and auxin. In GOA.

Response to abscisic acid; KEEP_AS_NON_CORE in curated review. In GOA.

Response to gibberellin; KEEP_AS_NON_CORE in curated review. Ethylene-gibberellin cross-talk. In GOA.

Negative regulation of ethylene-activated signaling pathway; ACCEPT in curated review. Core ETR1 function as negative regulator. In GOA. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0010105 resolves to 'negative regulation of ethylene-activated signaling pathway', not 'negative regulation of ethylene-activated'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Protein serine/threonine kinase activity; ACCEPT in curated review. Core NEK3 kinase function. In GOA.

ATP binding; ACCEPT in curated review. NEK3 kinase domain binds ATP. In GOA.

Protein binding; in GOA. NEK3 interacts with substrates. However generic.

Microtubule bundle formation; REMOVE in curated review for BRI1 but similar reasoning applies. For NEK3 specifically, the curated review proposes GO:0005874 (microtubule) as the correct localization. Microtubule bundle formation may not be the correct function. Annotated REMOVE in the context of BRI1 review.

Pollen development; curated review includes this as relevant to NEK3 function. Consistent with NIMA-kinase roles in cell division during gametophyte development. In GOA.

Cortical microtubule; consistent with NEK3 role in microtubule organization. In GOA.

Cytoskeleton; NEK3 associates with cytoskeletal structures. In GOA.

Cytoplasm; NEK3 is cytoplasmic. In GOA.

Correct prediction but already well-established in curated annotations (accepted in ai-review.yaml). AmyE is the archetypal bacterial alpha-amylase with EC 3.2.1.1 assignment, crystal structure (PDB:1BAG), and IEA/IDA evidence in GOA. The model correctly identifies the core molecular function.

Correct prediction and well-supported biologically. AmyE catalyzes endohydrolysis of starch alpha-1,4-glucosidic linkages. This biological process is consistent with the curated review (MODIFY suggestion toward more specific polysaccharide catabolic process term, but starch catabolic process is appropriate). Already in training data via IEA annotations.

Correct prediction for cellular component. AmyE is a secreted enzyme exported via the Sec pathway and functions in the extracellular environment. This is well-established and present in curated GOA annotations (accepted in ai-review.yaml).

Correct and core molecular function of AprE (Subtilisin E), which contains the canonical Asp-His-Ser catalytic triad (Asp138-His170-Ser327) of the peptidase S8 family. Already in curated annotations (IBA evidence, accepted in ai-review.yaml). This is already in training data.

Generic protein binding is uninformative and represents a frequency-biased prediction common to many secreted proteases in training data. The GO curation guidelines explicitly discourage this term as it conveys no specific functional information. AprE's relevant interactions (propeptide chaperone, substrate binding) are captured by more specific terms.

Correct biological process prediction. AprE performs extracellular proteolysis for nutrient acquisition and for processing signaling peptide precursors (PhrC/CSF). Already in curated annotations (IBA evidence, accepted in ai-review.yaml). In training data.

Incorrect prediction. Subtilisin E is a bacterial secreted protease from B. subtilis with no physiological role in blood coagulation. While subtilisins can cleave some coagulation factors in vitro (due to broad serine protease specificity), there is no evidence that AprE functions in blood coagulation in vivo. This is a clear nonparalog incorrect prediction, likely confusing AprE with mammalian serine proteases like thrombin or plasmin that have roles in coagulation. AprE functions in nutrient acquisition and quorum sensing pheromone processing in B. subtilis.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0005615 is 'obsolete extracellular space'. The raw and canonical labels denote the same biological concept. Ontology status is flagged separately, so the biological novelty and precision assessment is retained. The pair is classified CNN; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: Correct cellular component prediction. AprE is a secreted enzyme exported via the Sec pathway; UniProt states 'Secreted'. Already present in curated annotations (extracellular region accepted in ai-review.yaml). In training data. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0005615 resolves to 'obsolete extracellular space', not 'extracellular space'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Generic protein binding is uninformative and a frequency-biased prediction common for transcription factors in training data. ComK's relevant interactions (MecA adaptor, ClpC/ClpP protease) are captured by more specific terms. Per GO curation guidelines, this term should be avoided.

Incorrect prediction. ComK is a DNA-binding transcription activator (master regulator of competence) with no known ATPase or ATP-binding activity. ComK binds K-box DNA elements as a tetramer. The ATP binding prediction likely reflects contamination from other competence-related proteins (e.g., ComA, the ATP-binding type IV secretion machinery) or from the MecA-ClpC-ClpP proteolytic system that degrades ComK. ComK's protein sequence contains no nucleotide-binding motifs. An independent OpenScientist investigation (see file reference) confirms this NPI across five evidence types (sequence, domain family, curated database, cryo-EM structure, mechanism) -- the 192-aa ComK has no Walker A P-loop motif and no ATP-binding motif or nucleotide-binding fold; it is a sequence-specific DNA-binding transcription factor that activates transcription by stabilizing RNA polymerase, with no ATP step.

Incorrect prediction. ComK has no ATPase activity. As a DNA-binding transcription factor that activates late competence genes, ComK does not couple ATP hydrolysis to any molecular function. The ATPase activity prediction likely reflects training data contamination from competence-related ATP-consuming machinery (e.g., ComFA helicase, DNA uptake systems) present in the same biological context as comK. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0016887 resolves to 'ATP hydrolysis activity', not 'ATPase activity'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Correct novel prediction. ComK is known to form a dimer-of-dimers (tetramer) that binds K-box DNA elements at competence gene promoters. Self-association via homooligomerization is required for its function as a transcriptional activator. This is not currently in the GOA curated annotations for comK, making this a correct novel annotation. The deep research confirms ComK forms a dimer-of-dimers to bind K-box elements.

Correct prediction of a core biological process for ComK. ComK is the master transcriptional regulator of genetic competence and is specifically required for DNA-mediated transformation in B. subtilis. The GOA annotations include competence-related terms and this is well-established in curated annotations. In training data.

Incorrect prediction. ComK is the master regulator of genetic competence, not sporulation. ComK-expressing cells are in the K-state (competent state) which is mutually exclusive with sporulation commitment. ComK does not localize to the endospore core and has no known role in sporulation. This is a context confusion, likely because competence and sporulation are both stationary-phase responses in B. subtilis and share some regulators (e.g., Spo0A~P relieves AbrB repression of both pathways). [ONTOLOGY_LABEL_AUDIT] GO:0042606 does not resolve in the configured current GO ontology. The raw model label 'endospore core' is retained for provenance and the unresolved identifier is explicitly flagged.

Current-snapshot audit reclassified COR to CNN because the exact GO ID is already present in current local GOA. This exact-join decision supersedes conflicting conclusions in the earlier rationale. Retained biological rationale. Correct novel prediction. DivIVA forms oligomeric structures through its N-terminal membrane-binding domain and C-terminal coiled-coil regions that mediate tetramerization and higher-order self-assembly. DivIVA homooligomerization is essential for its function as a scaffold at cell poles and septa. This self-association is not currently annotated in the GOA curated set for divIVA, making it a correct novel prediction.

Incorrect prediction. DivIVA is a cell division scaffold protein that senses membrane curvature at cell poles and septa to recruit the MinJ/MinCD system for division site selection. DivIVA has no role in flagellum assembly. This is a major functional misassignment, likely resulting from confusion with other membrane-associated proteins involved in cell polarity. The error may arise from training data co-occurrence of cell division proteins with flagellar proteins, both being cell-shape-related processes.

Incorrect prediction. DivIVA has no role in flagellum-dependent swarming motility. DivIVA is a division plane organizer that functions through the MinCD system, not through flagellar regulation. This is a downstream consequence of the incorrect flagellum assembly prediction above.

While technically correct (DivIVA forms oligomeric complexes at poles and septa, interacts with MinJ/MinCD), this term is too generic. The curated annotation already has cytoplasm/membrane localizations. More specific terms would include the DivIVA-MinJ complex or division septum. The prediction captures a real aspect of DivIVA biology but at an overly broad level.

Correct and core molecular function of FliY. FliY contains a CheC-like phosphatase domain and has been directly demonstrated to catalyze CheY-P dephosphorylation in vitro (PMID:12920116). Already in curated annotations (IBA evidence, accepted in ai-review.yaml). In training data.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0071977 is 'bacterial-type flagellum-dependent swimming motility'. The canonical GO term is supported and absent as an exact ID from the frozen GOA. The pair is classified COR; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: Correct prediction. FliY regulates swimming motility by dephosphorylating CheY-P at the flagellar motor, controlling the duration of chemotactic signals. This is consistent with curated annotations (regulation of flagellum-dependent cell motility, accepted in ai-review.yaml). In training data. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0071977 resolves to 'bacterial-type flagellum-dependent swimming motility', not 'regulation of bacterial-type flagellum-dependent swimming motility'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Correct biological process. FliY catalyzes the dephosphorylation of phosphorylated CheY (CheY-P) to terminate chemotaxis signaling. The more specific term in training data captures this function. In training data.

Incorrect prediction. FliY is a flagellar switch/C-ring component and CheY-P phosphatase, not a flagellum assembly factor. FliY's C-terminal FliN-like domain integrates it into the pre-assembled flagellar switch complex (C-ring), but FliY does not have a role in assembling the flagellum from component parts. This confusion arises from FliY being a component of the flagellum rather than an assembler of it.

Partially correct but less precise. FliY primarily regulates swimming (individual flagellar rotation via CheY-P levels), not swarming motility per se. Swarming is a collective multicellular behavior with additional regulatory components. The curated annotation uses 'regulation of bacterial-type flagellum-dependent cell motility' which is more accurate. LSP because swarming is an oversimplification.

Incorrect prediction. FliY localizes to the flagellar C-ring (switch complex), not the cell division septum. There is no evidence that FliY is present at the septum. This is a clear misassignment, likely confusing FliY with other B. subtilis proteins like SpoIIE or FtsZ that have septum-related functions. An independent OpenScientist investigation (see file reference) confirms this NPI with high confidence -- FliY is localized in the flagellar switch complex (C-ring of the basal body), annotated to bacterial-type flagellum basal body and plasma membrane, with zero cell-division interactors. It also identifies a likely error origin -- FliY carries the "SpoA-like" fold label (SSF101801), where SpoA means Surface Presentation Of Antigens (a T3SS export-apparatus fold), not sporulation/septation; in Bacillus that token, plus the shared oligomeric-ring-at-the-membrane analogy with FtsZ, readily produces a spurious cell-septum call.

Partially correct. FliY is a component of the flagellar C-ring (switch complex), which is associated with the basal body. However, the more specific and accurate term would be the flagellar switch complex (GO:0009374) rather than the broader basal body. Still biologically reasonable as FliY is a flagellar component.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0009374 is 'biotin binding'. The canonical GO ID denotes a different, unsupported function from the raw label. The pair is classified NPI; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: The raw label says flagellar switch complex, but GO:0009374 is biotin binding. FliY is a flagellar switch phosphatase with no established biotin binding activity. The malformed pair is therefore NPI by its GO ID. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0009374 resolves to 'biotin binding', not 'flagellar switch complex'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Correct prediction. FliY is a cytoplasmic protein that localizes to the flagellar C-ring at the cytoplasm-membrane interface. Cytoplasmic localization is consistent with curated annotations. In training data.

Correct and core molecular function of FtsZ. FtsZ is a tubulin homolog with intrinsic GTPase activity that drives polymerization dynamics and Z-ring treadmilling. Supported by crystal structures and biochemical data. Already in curated GOA (IBA evidence, accepted in ai-review.yaml). In training data.

Correct prediction. FtsZ polymerizes into protofilaments through head-to-tail GTP-dependent self-association, forming the Z-ring. Self-association/identical protein binding is a fundamental molecular function of FtsZ. Already present in curated annotations (IBA evidence, accepted). In training data.

Correct and core biological process for FtsZ. FtsZ is the essential cell division protein that forms the Z-ring scaffold organizing the divisome for cytokinesis in bacteria. Already in curated annotations (IBA evidence, accepted). In training data.

Correct cellular component prediction. FtsZ assembles at the future division site to form the Z-ring, which defines the location of the division septum. FtsZ is the defining component of the division septum machinery. In training data.

Correct and highly specific prediction. The Z ring is the defining structure formed by FtsZ at the cell division site. FtsZ is the eponymous component of the Z ring. Already in curated annotations (accepted). In training data. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0032153 resolves to 'cell division site', not 'Z ring'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Generic protein binding is uninformative and a frequency-biased prediction for this translocase. SecA interacts with preproteins, SecYEG, and SecDF through highly specific interactions, none of which are captured by the generic protein binding term. Per GO curation guidelines, this term is discouraged.

Correct and core biological process for SecA. SecA is the ATPase motor of the Sec translocase that drives post-translational protein translocation across the cytoplasmic membrane. This is SecA's defining function. Already in curated annotations (accepted in ai-review.yaml). In training data.

Incorrect prediction. SecA has no known role in regulation of transcription. SecA is the motor ATPase of the Sec protein secretion system and functions in post-translational protein translocation, not transcription regulation. This prediction appears to be a spurious association, possibly from training data where secA appears in regulatory contexts (e.g., secA expression is autoregulated) but SecA protein itself does not regulate transcription.

Incorrect prediction. SecA is a bacterial protein localized to the cytoplasm and membrane of B. subtilis itself, not to the cytoplasm of a host cell. SecA is not a virulence factor or pathogen effector - B. subtilis is not a pathogen. This annotation is completely inappropriate for a housekeeping protein secretion ATPase and reflects a clear training data error, likely confusing B. subtilis SecA with pathogen-derived proteins that localize to host cells.

Correct prediction: sigF is a well-characterized sigma-70 family sigma factor. This term is already in GOA with IBA evidence (GO_REF:0000033) and IEA evidence (GO_REF:0000120), and is accepted as the primary molecular function in the curated AIGR review. The SFT model correctly identifies the core MF of this protein.

Frequency-biased prediction. While sigF does bind proteins (SpoIIAB anti-sigma factor, RNAP core subunits), GO:0005515 (protein binding) is explicitly discouraged in GO curation guidelines as it is uninformative without specifying the binding partner or functional context. The curated AIGR review flags this term as one to avoid. SigF protein-protein interactions are better captured by GO:1903865 (sigma factor antagonist complex), which is in GOA. This prediction likely reflects training data frequency bias toward common generic MF terms rather than protein-specific function.

Uncertain. SigF is activated during sporulation, which can be triggered by nutrient stress and other environmental signals. However, sigF is not characteristically described as a stress-response sigma factor -- that role in B. subtilis belongs to SigB. SigF is the forespore-early sigma factor activated after asymmetric septation as part of a developmental program, not a general stress response regulator. The curated GOA and AIGR review do not include this term; the accepted BP terms are regulation of DNA-templated transcription (GO:0006355) and sporulation (GO:0030435). While sporulation onset is stress-linked, the mechanistic activation of sigF (partner-switching via SpoIIE/SpoIIAA/SpoIIAB) is developmental rather than a canonical stress response. Cannot fully refute but this annotation would be misleading given sigF's well-characterized forespore-specific role. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0043620 resolves to 'obsolete regulation of DNA-templated transcription in response to stress', not 'regulation of DNA-templated transcription in response to stress'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Correct prediction: sigF plays an essential role in sporulation and is already annotated with GO:0030435 via IEA evidence (GO_REF:0000043, UniProtKB-KW:KW-0749) in GOA. The curated AIGR review accepts this annotation; sigF loss-of-function mutations abolish spore formation. The SFT model correctly identifies this biological process.

Less specific than existing annotation. sigF is annotated in GOA with GO:1903865 (sigma factor antagonist complex, IBA evidence), which is a child term of protein-containing complex. The curated AIGR review accepts the more specific sigma factor antagonist complex annotation, reflecting the well-characterized SigF-SpoIIAB inhibitory complex that is central to the partner-switching regulatory mechanism. GO:0032991 is a trivially broad term; the model should have predicted the more specific GO:1903865.

Incorrect. sigF encodes a cytoplasmic sigma factor that associates with the RNA polymerase core enzyme in the cytoplasm. It is not a membrane protein and is not localized to the plasma membrane. UniProt does not annotate sigF with any membrane localization. The RL review similarly flagged membrane-associated CC predictions as incorrect. This error likely reflects the model confusing the sigma factor itself with membrane-associated sporulation proteins (e.g., SpoIIE, which is a septal membrane protein that regulates SigF but is not SigF itself), or conflating sigF with the downstream targets of the SigF regulon which include membrane-associated proteins.

Incorrect. sigF is a cytoplasmic sigma factor (transcription initiation factor) and has no structural role in the spore wall. GO:0031160 (spore wall) is an appropriate annotation for structural spore wall proteins (e.g., CotA, CotB, SpoVID), not for the transcription factors that regulate their expression. The RL review likewise flagged spore wall as an incorrect CC prediction. This represents a category-level error: the model predicts the cellular compartment associated with sporulation rather than the compartment where the transcription factor itself resides and functions.

Generic protein binding is uninformative and a frequency-biased prediction. SigG's relevant interactions (core RNA polymerase, anti-sigma factor Gin/CsfB) are captured by more specific terms. Per GO curation guidelines, this term is discouraged.

Too generic for a well-characterized sigma factor. While SigG responds to developmental signals (forespore compartmentalization, anti-sigma factor Gin activity), this high-level term provides no useful information. More specific terms like 'sporulation resulting in formation of a cellular spore' or 'sigma factor activity' are in curated annotations.

Too generic. Same issue as response to stimulus - correct at a high level but provides no useful information for this specific forespore sigma factor.

Too generic. Sporulation is triggered by starvation/stress, so this is not incorrect, but 'response to stress' is a high-level ancestor term that provides no insight into SigG's specific forespore sigma factor role.

Too generic. SigG regulates transcription of forespore genes, which encompasses metabolic processes, but this term is too broad to be informative. All sigma factors could be annotated with this term.

Too generic ancestor term. This and all subsequent regulation of transcription ancestor terms provide progressively less information than 'regulation of DNA-templated transcription' which is the relevant curated annotation.

Too generic ancestor term of regulation of transcription.

Too generic ancestor term of regulation of transcription.

Too generic. Sporulation is a response to nutritional stress, but this term applies to all stress-responsive genes. SigG specifically drives late forespore gene expression, which is far more specific.

Too generic ancestor term of regulation of transcription.

Too generic ancestor term. Transcription regulation is a nitrogen compound metabolic process but this is too broad for SigG.

Too generic ancestor term of regulation of transcription.

Too generic ancestor term of regulation of DNA-templated transcription. Getting closer but still not specific to forespore sigma factor biology.

Too generic ancestor term of regulation of transcription.

Too generic. While correct, this is an ancestor of the curated annotation 'regulation of DNA-templated transcription'. Sigma factors regulate gene expression but this term applies to all regulators.

Correct and reasonably specific prediction. SigG regulates forespore transcription in response to developmental/stress signals (starvation triggering sporulation). The 'in response to stress' aspect reflects sporulation being a stress response. This is in the same part of the GO hierarchy as curated annotations and supported by the biology. In training data.

Too generic ancestor term of regulation of transcription.

Too generic ancestor term of regulation of DNA-templated transcription.

Correct prediction at the appropriate level of specificity. SigG regulates transcription of late forespore genes (e.g., sspA, sspB, spoVT) by directing RNA polymerase to sigG-dependent promoters. This is in the curated annotation set (IBA evidence, accepted in ai-review.yaml). In training data.

Slightly less specific ancestor term than regulation of DNA-templated transcription. The DNA-specific term is more appropriate for a sigma factor.

Too generic. SigG is part of the RNA polymerase holoenzyme complex during sporulation but 'protein-containing complex' is an ancestor term of the relevant complex. The specific RNA polymerase sigma-G holoenzyme complex would be the appropriate term.

Correct cellular component prediction. SigG is a subunit of the RNA polymerase holoenzyme complex, which is a transcription regulator complex. This is consistent with curated annotations for sigma factors. In training data.

Generic protein binding is uninformative and a frequency-biased prediction. SigK's relevant interactions (core RNA polymerase, GerE transcription factor) are captured by more specific terms. Per GO curation guidelines, this term is discouraged.

Incorrect prediction. SigK is a transcriptional activator (sigma factor) that directs RNA polymerase to mother-cell-specific promoters to activate late sporulation genes (coat proteins cotA, cotB, cotD, cotE, cotH; cortex and germination genes). Sigma factors are positive regulators of transcription, not negative regulators. The curated ai-review.yaml explicitly notes that annotations attributing negative regulation to sigma factors are incorrect. This confusion likely arises from the complex regulatory network around sigK where GerE represses some sigK-activated genes, but SigK itself is an activator.

Correct biological process prediction. SigK directs RNA polymerase to sigK-dependent promoters to transcribe late sporulation genes in the mother cell. DNA-templated transcription is the core biological process for a sigma factor. Already in curated annotations (accepted). In training data.

Correct and specific biological process. Sigma factors specifically function at the transcription initiation step by binding promoter elements (-10 and -35 regions) to recruit core RNA polymerase. SigK recognizes specific sigK-dependent promoters in mother cell genes. Already in curated annotations. In training data.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0009295 is 'nucleoid'. The canonical GO term is not resolved by the current reference evidence. The pair is classified UNC; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: GO:0009295 is nucleoid, not bacterial-type RNA polymerase holoenzyme complex. SigK acts with RNA polymerase on chromosomal promoters, so nucleoid is biologically plausible but broad and indirect; LSP is retained for the GO ID while the two distinct raw labels are preserved and flagged. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0009295 resolves to 'nucleoid', not 'bacterial-type RNA polymerase holoenzyme complex'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0031381 is 'viral RNA-directed RNA polymerase complex'. The canonical GO term is taxonomically incompatible with this gene product. The pair is classified NPI; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: The raw label says bacterial RNA polymerase holoenzyme, but GO:0031381 is viral RNA-directed RNA polymerase complex. A Bacillus sporulation sigma factor is not part of a viral polymerase complex, so the pair is NPI. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0031381 resolves to 'viral RNA-directed RNA polymerase complex', not 'bacterial-type RNA polymerase holoenzyme complex'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0009295 is 'nucleoid'. The canonical GO term is not resolved by the current reference evidence. The pair is classified UNC; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: GO:0009295 is nucleoid, not bacterial-type RNA polymerase holoenzyme complex. SigK acts with RNA polymerase on chromosomal promoters, so nucleoid is biologically plausible but broad and indirect; LSP is retained for the GO ID while the two distinct raw labels are preserved and flagged. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0009295 resolves to 'nucleoid', not 'bacterial nucleoid'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Generic protein binding is uninformative and a frequency-biased prediction common for transcription factors. Spo0A's relevant interactions (Spo0B phosphotransfer, RNA polymerase) are captured by more specific terms. Per GO curation guidelines, this term is discouraged.

Correct and core molecular function. Spo0A is the master transcriptional regulator of sporulation initiation, functioning as both an activator and repressor by binding 0A-box sequences (5'-TGNCGAA-3'). Already in curated annotations (IEA evidence, accepted in ai-review.yaml). In training data.

Correct prediction. Spo0A~P binds specifically to 0A-box sequences (5'-TGNCGAA-3') in target promoters. The crystal structure (PDB 1LQ1) confirms sequence-specific DNA binding by the C-terminal HTH domain. Already in curated annotations (proposed modification in ai-review.yaml). In training data.

Correct and highly specific prediction. Spo0A is the terminal response regulator of the KinA/KinB -> Spo0F -> Spo0B -> Spo0A phosphorelay. The N-terminal CheY-like receiver domain accepts phosphoryl groups on Asp-56. This is a core curated function (accepted in ai-review.yaml). In training data.

Correct prediction. Spo0A at low/moderate phosphorylation levels activates biofilm formation programs in B. subtilis, including pellicle and colony biofilm development. This is well-established in the literature and present in curated annotations (ai-review.yaml). In training data.

Correct prediction. High-level Spo0A~P is essential for asymmetric cell division during sporulation initiation - it activates spoIIA, spoIIE, and spoIIG operons required for polar septum formation. Already in curated annotations. In training data.

Correct and highly specific prediction. Spo0A is the master regulator that commits cells to sporulation when phosphorylated to high levels. It directly activates early sporulation operons. This is the defining biological role of Spo0A and is in curated annotations. In training data.

Correct prediction. Spo0A~P activates expression of spoIIE and other genes required for the asymmetric polar septum that initiates sporulation. Spo0A is required for polar septum assembly during sporulation. In curated annotations. In training data.

Correct prediction. Spo0A~P functions as a transcriptional activator for numerous sporulation genes (spoIIA, spoIIG, spoIIE operons). This is a core transcriptional function of Spo0A. In curated annotations (accepted in ai-review.yaml). In training data.

Correct cellular component. Spo0A is a cytoplasmic response regulator that translocates to the nucleoid upon phosphorylation. The cytoplasmic localization is appropriate. In curated annotations. In training data.

Correct cellular component. Spo0A~P binds to 0A-box sequences in promoter regions of chromosomal DNA, thus associating with the bacterial nucleoid during its regulatory function. In curated annotations. In training data.

Generic protein binding is uninformative and a frequency-biased prediction. SpoIIE's relevant interactions (SpoIIAA-P substrate, FtsZ, DivIVA) are captured by more specific terms. Per GO curation guidelines, this term is discouraged.

Incorrect prediction. SpoIIE is a PP2C-family sporulation-specific phosphatase that dephosphorylates SpoIIAA-P to activate sigma-F in the forespore. SpoIIE has absolutely no role in flagellum assembly. This is a major functional misassignment. The error likely arises because SpoIIE has a large N-terminal transmembrane domain with multiple helices that superficially resembles transmembrane regions of some flagellar proteins, or from training data co-occurrence artifacts with other membrane proteins in sporulation-related literature.

Incorrect prediction. SpoIIE has no role in flagellum-dependent motility. This is a downstream consequence of the incorrect flagellum assembly prediction. SpoIIE functions exclusively in sporulation compartmentalization, not in motility.

Correct cellular component prediction. SpoIIE localizes to the polar sporulation septum and becomes enriched on the forespore-facing membrane where it activates sigma-F in a compartment-specific manner. SpoIIE is retained in the forespore compartment. The endospore-forming forespore is an appropriate cellular component. Already in curated annotations (accepted in ai-review.yaml). In training data.

Correct and core molecular function of SpoIIGA. SpoIIGA is a novel type of membrane-associated aspartic protease that cleaves pro-sigmaE to generate active sigma-E during sporulation. The catalytic aspartate D183 is essential for activity. Already in curated annotations (IEA evidence, accepted in ai-review.yaml). In training data.

[FROZEN_ONTOLOGY_ADJUDICATION] Assessed the supplied GO ID rather than the intended raw label: GO:0043621 is 'obsolete protein self-association'. The canonical GO ID denotes a different, unsupported function from the raw label. The pair is classified NPI; the raw source pair remains unchanged. Historical rationale about the intended raw label is retained below for provenance, not as support for the canonical GO ID: Correct novel prediction. SpoIIGA has been shown to self-associate in the membrane (PMID:18378688 showed SpoIIGA expressed in E. coli is membrane-associated and self-associated after detergent treatment). The dimeric active site architecture resembles retroviral proteases which form homodimers. This self-association property is not currently in the GOA annotations for spoIIGA and represents a correct novel prediction. [ONTOLOGY_LABEL_AUDIT] Raw model pair mismatch: GO:0043621 resolves to 'obsolete protein self-association', not 'protein self-association'. The raw ID and label are retained for provenance; the assessment applies to the GO ID and explicitly flags the label error.

Correct biological process. SpoIIGA proteolytically processes the N-terminal pro-sequence of pro-sigmaE to release mature sigmaE. Proteolysis is the core biological process. Already in curated annotations (accepted in ai-review.yaml). In training data.

Correct cellular component. SpoIIGA is an integral membrane protein with multiple N-terminal transmembrane segments anchored to the sporulation septum (a specialized domain of the plasma membrane). The plasma membrane annotation is correct and already in curated annotations (accepted in ai-review.yaml). In training data.

Dph2 forms a heterodimeric complex with Dph1 and interacts with eEF2. While protein binding is a generic term not annotated in GOA for dph2, it is implicit in the complex biology. However, this is an over-generic prediction; the correct annotation would be more specific complex or interaction terms. CNN because the Dph1-Dph2 complex interaction is in GOA training data via the complex annotation (GO:0120513). Assessed CNN with note that more specific terms are preferred.

ACCEPT in curated review. Dph2 is part of the Dph1-Dph2 heterodimeric complex that catalyzes the first step of diphthamide biosynthesis, transferring a 3-amino-3-carboxypropyl group from SAM to His715 of eEF2. GO:0017183 (protein histidyl modification to diphthamide) is in GOA for dph2 via both IBA and IEA. CNN because it is already in training data.

Diphthamide biosynthesis is required for accurate translation and thus broadly affects development. Cartilage development is not a documented specific function of dph2 in zebrafish; the curated review does not annotate dph2 to cartilage development. DPH2 loss in humans causes a developmental disorder with skeletal involvement, but cartilage-specific function has not been mechanistically established for dph2 in zebrafish. UNC because the link is plausible but not substantiated.

Dph2 is a cytoplasmic enzyme that modifies cytoplasmic eEF2. The Dph1-Dph2 complex is localized to the cytoplasm where translation occurs. GO:0005737 is consistent with dph2 function. While cytoplasm is not explicitly listed in GOA for dph2, the complex annotation (GO:0120513) implies cytoplasmic localization. CNN because cytoplasmic localization is inferred by the training data context.

Dph2 is a cytoplasmic enzyme that modifies cytoplasmic eEF2 (translation elongation factor 2). It is not annotated to the nucleus in GOA, and there is no evidence for nuclear localization of the Dph1-Dph2 complex. The curated review does not annotate dph2 to nucleus. This prediction is incorrect; the enzyme acts on a cytoplasmic substrate and has no documented nuclear function.

ACCEPT in curated review. 5'-flap endonuclease activity is the defining molecular function of FEN1. The enzyme recognizes flap structures and makes a single endonucleolytic cut at the flap base. GO:0017108 is in GOA for fen1 via both IBA and IEA annotations. CNN because it is already in training data.

KEEP_AS_NON_CORE in curated review. Zebrafish fen1 insertional mutants display severely defective retinal development with reduced and disorganized retinal neurons (PMID:15716491). This is a direct IMP annotation in GOA from a zebrafish mutagenesis screen reflecting the high proliferative demand of retinal progenitors. CNN because GO:0060041 is already in GOA training data for fen1.

FEN1 is essential for DNA replication, specifically for Okazaki fragment maturation during lagging-strand synthesis. GO:0006260 (DNA replication) is a less specific parent term of GO:0033567 (Okazaki fragment processing) and GO:0043137 (DNA replication, removal of RNA primer), both of which are in GOA for fen1. The broader DNA replication term is accurate but less informative. LSP because more specific terms are annotated in GOA.

ACCEPT in curated review. FEN1 processes Okazaki fragment flaps generated during lagging-strand DNA replication, cleaving displaced 5'-flaps to produce ligatable nicks. GO:0043137 (DNA replication, removal of RNA primer) covers this function and is in GOA for fen1. GO:0033567 is closely related and is in training data context. CNN because Okazaki fragment processing is a core documented function in GOA.

ACCEPT in curated review. FEN1 plays a critical role in long-patch base excision repair (LP-BER), trimming 5'-flap structures generated by strand displacement synthesis. GO:0006284 is in GOA for fen1 via IEA. CNN because it is already in training data.

FEN1 has been proposed to function in homologous recombination, processing intermediates such as Holliday junctions, but this is not its primary or well-established function. The curated review does not annotate fen1 to HR. GO:0000724 is not in GOA for fen1. UNC because there is limited evidence for a direct HR role in vivo.

FEN1 is not involved in nonhomologous end joining (NHEJ). NHEJ repairs DNA double-strand breaks with minimal resection, whereas FEN1 is a 5'-flap endonuclease that acts on 5'-flap structures in replication and base excision repair contexts. The curated review does not annotate fen1 to NHEJ, and GO:0006303 is not in GOA for fen1. This is an incorrect prediction confusing the types of DNA repair.

GO:0006310 (DNA recombination) is a broad ancestral term. FEN1 does not have a primary documented role in DNA recombination; its established biological processes are DNA replication and base excision repair. If any recombination role exists, more specific terms would apply. LSP relative to the specific repair/replication terms already annotated.

ACCEPT in curated review. FEN1 is a nuclear protein that localizes to the nucleolus (primary location under normal conditions) and nucleoplasm (upon DNA damage). GO:0005634 is in GOA for fen1 via both IBA and IEA. CNN because it is already in training data.

ufsp2 is a UFM1-specific cysteine protease (deUFMylase). GO:0019783 (ubiquitin-like protein peptidase activity) is a parent term of the more specific GO:0071567 (deUFMylase activity), which is in GOA for ufsp2 via IBA, ISS, and IEA. The model predicted a less specific term; the correct more specific term is already annotated. LSP because GO:0071567 is available and annotated.

ufsp2 is a deUFMylase that cleaves UFM1 from conjugated substrates and processes UFM1 precursors. It functions in ribosome disassembly and ER-associated ribosome quality control. Gravitaxis (orientation behavior in response to gravity) is not a documented function of ufsp2 and is not in GOA for ufsp2. This is an incorrect prediction likely arising from confusion with other zebrafish proteins (e.g. cryptochrome CRY which is annotated to gravitaxis). No evidence links ufsp2 to gravitaxis behavior.

ufsp2 is a deUFMylase in the UFM1 conjugation pathway, functioning in ER-associated ribosome quality control. Response to light stimulus is not a documented function of ufsp2. GO:0009416 is not in GOA for ufsp2. This prediction appears to be a cross-contamination from light-responsive proteins (such as cryptochromes or other zebrafish proteins). No biological connection between UFM1 deconjugation and light responses is established.

ufsp2 localizes to multiple compartments including cytoplasm, ER membrane, and nucleus. GO:0005737 (cytoplasm) is in GOA for ufsp2 via both ISS and IEA annotations. CNN because cytoplasm localization is in training data. Note that ER (GO:0005783) is the primary functional location for ufsp2 where it deUFMylates RPL26 on ER-bound ribosomes.

KEEP_AS_NON_CORE in curated review. MlcD retains one functional EF-hand with low-affinity calcium binding (Kd 52 uM), well above physiological levels. GO:0005509 is in GOA for mlcD via IBA and IDA. The protein cannot function as a physiological calcium sensor, but calcium binding is documented. CNN because the term is already in training data.

MODIFY in curated review (replace with GO:0032037 myosin I heavy chain binding). MlcD specifically binds the MyoD (Myo1D) heavy chain IQ motifs in a calcium-insensitive manner. GO:0032036 is in GOA for mlcD via IPI (PMID:12826013). CNN because the term is in training data, though the more specific child term is preferred.

Negative regulation of asexual reproduction is not documented for mlcD. The protein functions as the light chain of MyoD (Myo1D) at macropinocytic cups and actin waves. The curated review does not annotate mlcD to reproductive processes. UNC because the prediction is not refuted but lacks supporting evidence.

Aggregation involved in sorocarp development (Dictyostelium social behavior) is not documented for mlcD specifically. The curated review focuses on actin/membrane dynamics functions. While MyoD/MlcD may contribute to aggregation indirectly through actin dynamics, no direct evidence links mlcD to sorocarp development. UNC.

Regulation of ATP-dependent activity is not a documented function of mlcD. MlcD is a structural light chain that binds MyoD IQ motifs; it does not regulate ATP-dependent activity in a broader sense. The curated review removes the calmodulin-like enzyme regulator activity (GO:0030234) annotation for similar reasons. Not in GOA for mlcD. NPI.

Mitotic cytokinesis is not a documented function of mlcD. MlcD specifically associates with class I myosin MyoD (Myo1D) which localizes to macropinocytic cups and actin waves, not the contractile ring of cytokinesis (which is driven by myosin II). The curated review does not annotate mlcD to cytokinesis. NPI. An independent OpenScientist investigation (see file reference) confirms this and the other three disputed myosin-II/cytokinesis terms on mlcD (GO:0030837, GO:0043520, GO:0071976) -- mlcD is the dedicated essential light chain of MyoD, a single-headed class-I (unconventional) myosin, not a component of the conventional myosin-II (mhcA) contractile machinery -- so the myosin-II-filament-assembly and cytokinesis roles are a paralog/class misassignment.

Plasma membrane repair is not specifically documented for mlcD. While MyoD/MlcD is associated with membrane dynamics (macropinocytic cups), direct evidence for plasma membrane repair is absent. UNC because the cellular context is related to membrane dynamics but the specific process has not been shown.

Positive regulation of protein binding is not a documented function of mlcD. This is a very generic regulatory term without evidence for mlcD. The curated review does not annotate this function, and it is not in GOA for mlcD. NPI.

Negative regulation of actin filament polymerization is not documented for mlcD. MlcD is a myosin light chain that associates with the motor domain context of MyoD; it does not act as a regulator of actin polymerization. The curated review annotates mlcD to actin wave localization but not to regulatory functions in actin dynamics. NPI.

MlcD is a dedicated light chain for MyoD (Myo1D), a class I myosin. Regulation of myosin II filament assembly is not relevant to a class I myosin light chain. Myosin II and myosin I use different light chains and have fundamentally different mechanisms. The curated review does not annotate mlcD to myosin II function. NPI.

Regulation of GTPase activity is not documented for mlcD. This function is not characteristic of myosin light chains in general, and MlcD specifically is a structural calmodulin-like light chain for MyoD. Not in GOA for mlcD. NPI.

Nucleus organization is not a function of mlcD. MlcD is a cytoplasmic myosin light chain that localizes to actin structures (macropinocytic cups, actin waves). Not in GOA for mlcD. NPI.

Protein localization to cell division site is not documented for mlcD. MlcD localizes to macropinocytic cups and actin waves via its interaction with MyoD, not to division sites. Class I myosin light chains do not function at the cytokinetic ring. NPI.

Pinocytosis is supported by the localization of MlcD (via MyoD/Myo1D) to macropinocytic protrusions, cups, and freshly internalized vesicles (PMID:31774725). The curated review accepts actin wave and macropinocytic cup annotations for mlcD. While pinocytosis per se is not in GOA, the macropinocytic cup cytoskeleton annotation implies pinocytic function. CNN based on the curated review context; the training data includes the macropinocytic localization.

Phagocytosis engulfment is not specifically documented for mlcD/MyoD. MyoD localizes to macropinocytic cups but direct evidence for phagocytic cup involvement is limited. The curated review focuses on pinocytosis/macropinocytosis rather than phagocytosis specifically. UNC.

Actin filament polymerization is not a function of mlcD. MlcD is a calmodulin-like light chain for MyoD; it participates in actin-based motility via the motor, but does not promote actin polymerization. Not in GOA for mlcD. NPI.

MODIFY in curated review (more specific myosin I complex preferred). MlcD is part of the MyoD (class I myosin) complex, co-purifying as two copies per MyoD heavy chain. GO:0016459 (myosin complex) is in GOA for mlcD via IDA (PMID:12826013). CNN because the term is in training data.

ACCEPT in curated review (as macropinocytic cup cytoskeleton GO:0070687). MlcD localizes to macropinocytic cups via its interaction with MyoD/Myo1D (PMID:31774725). GO:0070685 (macropinocytic cup) is a closely related term. CNN because macropinocytic cup localization is in the training data context.

Cortical actin cytoskeleton is not explicitly annotated for mlcD in GOA. MlcD localizes to actin waves (GO:0062201) and macropinocytic cups, which are cortical actin structures. The cortical actin cytoskeleton annotation is plausible but not directly in GOA. UNC because it is consistent with the biology but not confirmed.

Nip7 interacts with Nop8p and the exosome subunit Rrp43p in yeast, and the PUA domain mediates RNA binding while the pre-PUA domain mediates protein interactions. While protein binding per se is not explicitly in GOA for nip7, it is implicit in the preribosome complex annotation (IBA). CNN because protein interactions are established in the training data context, though the prediction is too generic.

ACCEPT in curated context. Nip7 has a PUA (pseudouridine synthase and archaeosine transglycosylase) domain that binds polyuridine RNA. GO:0003723 (RNA binding) is in GOA for nip7 via IEA from InterPro PUA domain annotations. CNN because the term is already in training data.

GO:0042254 (ribosome biogenesis) is a parent term of GO:0042273 (ribosomal large subunit biogenesis), which is the more specific and accurate annotation for nip7 function. Nip7 specifically functions in 60S ribosomal subunit biogenesis (processing of 27S pre-rRNA). GO:0042273 is in GOA via IBA and ISS; GO:0042255 (ribosome assembly) also in GOA via IEA. LSP because more specific terms are annotated.

ACCEPT in curated review. Nip7 is a conserved trans-acting factor required for 27S pre-rRNA processing and 60S ribosomal large subunit biogenesis. GO:0042273 is in GOA for nip7 via IBA and ISS annotations. CNN because it is already in training data.

Nip7 is a ribosome biogenesis factor with no documented role in DNA damage response. The curated review does not annotate nip7 to DNA damage response, and GO:0006974 is not in GOA for nip7. However, ribosome biogenesis factors can be downregulated during DNA damage as part of growth arrest, making an indirect connection plausible. UNC because there is no direct evidence but cannot be firmly refuted.

Response to UV is not a documented function of nip7. Nip7 is a ribosome biogenesis factor that functions in 60S subunit assembly. GO:0009411 is not in GOA for nip7. UV response involves DNA repair and stress response factors, not ribosome biogenesis proteins specifically. NPI because there is no evidence for this function and it is biologically inconsistent with nip7 as a ribosome assembly factor.

ACCEPT in curated review. Nip7 is a nucleolar protein required for pre-rRNA processing in the nucleolus. GO:0005730 is in GOA for nip7 via IBA and IEA. Yeast Nip7p-GFP fusion localizes to the nucleolus. CNN because the annotation is already in training data.

N-acylsphingosine amidohydrolase (ceramidase) activity is incorrect for tlcd4b. Recent characterization of TLCD family members (Sheokand et al. 2025, PMID:39970228) established that non-CerS TLCD proteins (including TLCD1, TLCD2) are phospholipid acyltransferases (lysophospholipid acyltransferases), NOT ceramidases. Ceramidases hydrolyze ceramide to sphingosine and fatty acid, whereas TLCD4 subfamily proteins transfer acyl groups to lysophospholipids. The curated review proposes NEW annotation for lysophospholipid acyltransferase activity (GO:0071617) for tlcd4b. This is a fundamental mechanistic error confusing two different enzyme activities within the TLC domain superfamily. An independent OpenScientist investigation (see file reference) confirms this NPI -- tlcd4b belongs to the TLCD4 family (six of six InterPro signatures TLC-family, zero ceramidase-family) and contains none of the acid/neutral/alkaline ceramidase active-site constellations; Dictyostelium encodes dedicated ceramidases (dcd1/2/3) that are not tlcd4b, so there is no ceramidase role for it to fill.

Ceramide metabolic process is incorrect for tlcd4b. TLCD4 subfamily proteins function as phospholipid acyltransferases in phospholipid remodeling, not in ceramide metabolism. Ceramide metabolism is specific to ceramide synthases (CerS/LASS proteins) and ceramidases, which belong to different subfamilies of the TLC domain superfamily. The curated review annotates tlcd4b to phospholipid metabolic process rather than ceramide metabolic process. GO:0006672 is not in GOA for tlcd4b. NPI because ceramide metabolism is demonstrably incorrect based on TLCD family biochemistry. An independent OpenScientist investigation (see file reference) agrees tlcd4b is not a catalytic ceramide-metabolism enzyme; it rates any ceramide-process link as weak/indirect and non-catalytic at most, consistent with the review's more specific phospholipid-acyltransferase assignment (Sheokand et al. 2025). NPI as a catalytic/family misassignment.

ACCEPT in curated review (as MODIFY from GO:0016020 membrane). tlcd4b has six predicted transmembrane helices and is a multi-pass integral membrane protein. GO:0016021 is a more specific term than GO:0016020 (membrane), which is in GOA for tlcd4b. The curated review proposes modifying to GO:0016021 as more informative. CNN because membrane/integral membrane status is in the training data context.

Plasma membrane localization is incorrect for tlcd4b. TLCD family proteins, including TLCD1 and CerS proteins, localize to the endoplasmic reticulum (ER). The curated review accepts ER localization (GO:0005783) based on IBA annotation and TLC family biology. GO:0005886 is not in GOA for tlcd4b. Predicting plasma membrane for an ER-resident multi-pass transmembrane protein is an incorrect compartment assignment.

Extracellular matrix localization is incorrect for tlcd4b. tlcd4b is an integral membrane protein (six predicted TM helices) localized to the ER membrane. Multi-pass transmembrane proteins with an ER retention context are not secreted into the extracellular matrix. GO:0031012 is not in GOA for tlcd4b. This prediction is biologically impossible for a multi-pass ER membrane protein.

Too general for Akt1 characterization. BioReason-Pro predicts high-level GO hierarchy terms that provide no specific insight into Akt1 function. LSP (less specific than annotated core functions like protein serine/threonine kinase activity and insulin receptor signaling pathway).

Too general for Akt1 characterization. BioReason-Pro predicts high-level GO hierarchy terms that provide no specific insight into Akt1 function. LSP (less specific than annotated core functions like protein serine/threonine kinase activity and insulin receptor signaling pathway).

Too general for Akt1 characterization. BioReason-Pro predicts high-level GO hierarchy terms that provide no specific insight into Akt1 function. LSP (less specific than annotated core functions like protein serine/threonine kinase activity and insulin receptor signaling pathway).