id: P26266
gene_symbol: fepE
locus_tag: b0587
taxon:
  id: NCBITaxon:83333
  label: Escherichia coli K-12
status: COMPLETE
description: >-
  FepE DeepECTF prediction review. FepE was one of 12 proteins incorrectly predicted
  as histidine kinase (EC 2.7.13.3) by DeepECTF. This is a frequency-biased repetition
  error - none of the 12 predicted histidine kinases have sequence similarity to
  histidine kinase families, and 8 (including FepE) have experimentally validated
  different functions. FepE is a Wzz-family polysaccharide co-polymerase that regulates
  O-antigen chain length.
references:
  - id: PMID:37963869
    title: Functional annotation of enzyme-encoding genes using deep learning with transformer layers.
  - id: PMID:40703034
    title: Limitations of current machine learning models in predicting enzymatic functions for uncharacterized
      proteins.
  - id: PMID:12603743
    title: Regulation of Salmonella typhimurium lipopolysaccharide O antigen chain length is required
      for virulence; identification of FepE as a second Wzz.
  - id: PMID:22437828
    title: Structural characterization of closely related O-antigen lipopolysaccharide (LPS) chain
      length regulators.
  - id: file:ECOLI/fepE/fepE-hypotheses/function-hypothesis-go-0004713/openscientist.md
    title: OpenScientist hypothesis report for fepE protein tyrosine kinase activity
    publication_type: DEEP_RESEARCH
    findings:
    - statement: OpenScientist refuted kinase transfer to FepE because the protein has only a periplasmic
        Wzz/PCP module and lacks cytoplasmic kinase machinery.
      supporting_text: FepE possesses **only** that periplasmic Wzz domain and does **not** carry the
        cytoplasmic BY-kinase machinery.
source_documents:
  - genes/ECOLI/fepE/fepE-hypotheses/function-hypothesis-go-0004713/openscientist.md
  - publications/PMID_37963869.md
  - publications/PMID_40703034.md
  - publications/PMID_12603743.md
  - publications/PMID_22437828.md
predictions:
  - source_method: DeepECTF
    source_version: "2023"
    source_reference_id: PMID:37963869
    predicted_term:
      id: EC:2.7.13.3
      label: histidine kinase
    predicted_term_type: EC
    review:
      assessment: REP
      confidence_score: 0
      error_type: FREQUENCY_BIAS
      summary: >-
        Repetition/frequency bias error. FepE was one of 12 proteins incorrectly annotated
        as histidine kinase (EC 2.7.13.3) by DeepECTF. None of these 12 have sequence
        similarity to histidine kinase families, and 8 have been annotated with different
        experimentally validated functions. E. coli has 29 actual histidine kinases, making
        this a high-frequency label in the training data. The model appears to default to
        this common EC number when it cannot identify informative sequence features.
        FepE is actually a Wzz-family polysaccharide co-polymerase (PCP-1 class) that
        determines very-long modal O-antigen chain length. The OpenScientist kinase audit
        independently confirmed that FepE has the periplasmic Wzz domain rather than the
        cytoplasmic machinery required for a bacterial tyrosine kinase, strengthening the
        conclusion that no kinase domain is present. FepE functions through structural
        oligomerization, not catalysis.
      supported_by:
        - reference_id: PMID:40703034
          supporting_text: "out of the 12 proteins annotated as histidine kinase (EC 2.7.13.3) in the 453 DeepECTF predictions, none of them have sequence similarity to histidine kinase families, and 8 have been annotated with different and experimentally validated functions (such as ferric enterobactin transport protein FepE for b0587)"
        - reference_id: PMID:40703034
          supporting_text: "This type of error may be due to inherent limitations in how AI methods operate...the model is expected to make frequency-dependent predictions that reflect the training data, as demonstrated with histidine kinases"
        - reference_id: file:ECOLI/fepE/fepE-hypotheses/function-hypothesis-go-0004713/openscientist.md
          supporting_text: >-
            FepE's activity, process, and location are all inconsistent with GO:0004713:
            no ATP-binding catalytic domain (activity), no phosphotransfer role in its
            pathway (process), and its bulk faces the periplasm while the tiny cytoplasmic
            tail cannot host a kinase