ARBA00023137

id: ARBA00023137
description: 'Rule that applies UniProt keyword "Tyrosine-protein kinase" to proteins with various kinase-related InterPro domains, CATH FunFam families, and PANTHER families across diverse taxonomic groups. Contains 98 condition sets but provides no GO annotations.'
status: COMPLETE
rule_type: ARBA
rule:
  rule_id: ARBA00023137
  condition_sets: []
  go_annotations: []
  reviewed_protein_count: 0
  unreviewed_protein_count: 203241
  created_date: '2020-05-12'
  modified_date: '2025-05-15'
  entries: []
review_summary: 'This ARBA rule demonstrates severe design problems that make it unsuitable for automated protein annotation. With 98 condition sets incorporating 35 InterPro domains, 106 CATH FunFam families, and 4 PANTHER families, it violates basic principles of parsimony and maintainability. Deep research reveals that the rule attempts to capture fundamentally diverse kinase subfamilies under a single "tyrosine-protein kinase" keyword, which is biologically inaccurate for many captured proteins. The condition sets target receptor tyrosine kinases (appropriately restricted to metazoans), bacterial kinases, plant-specific kinases, and dual-specificity kinases - all of which require distinct functional annotations. Only 17 of 98 condition sets have taxonomic restrictions, leading to inappropriate cross-lineage annotations. The rule applies across 15 different taxonomic groups from bacteria to specific mammalian lineages without consistent biological rationale. Most critically, using keyword annotation instead of structured GO molecular function terms severely limits the utility for computational functional analysis. This represents a classic example of rule complexity arising from incremental additions without systematic design principles.'
action: DEPRECATE
action_rationale: 'This rule should be deprecated and replaced with a smaller set of well-designed rules. The current rule is overly complex (98 condition sets), provides minimal functional information (keyword only), and likely generates many false positives due to promiscuous domain matching. A better approach would be to create separate rules for distinct kinase subfamilies (e.g., receptor tyrosine kinases, non-receptor tyrosine kinases, dual-specificity kinases) with appropriate GO molecular function terms and proper taxonomic restrictions. The high number of condition sets makes this rule unmaintainable and impossible to validate systematically.'
suggested_modifications:
- 'Replace with separate rules for major kinase subfamilies'
- 'Use specific GO molecular function terms instead of generic keywords'
- 'Reduce condition sets to <12 per rule for maintainability'
- 'Apply appropriate taxonomic restrictions based on kinase evolution'
- 'Eliminate redundant condition sets that capture overlapping protein sets'
- 'Focus on high-specificity CATH FunFam families rather than broad InterPro families'
parsimony:
  assessment: OVERLY_COMPLEX
  notes: 'The rule contains 98 condition sets with 35 unique InterPro domains, 106 unique CATH FunFam families, and 4 PANTHER families, which far exceeds the recommended maximum of 12 for maintainable rules. Analysis is computationally infeasible due to the number of pairwise comparisons required (4,753 pairs). Deep research on kinase classification indicates this complexity arises from attempting to capture evolutionarily distinct kinase subfamilies (receptor tyrosine kinases, non-receptor kinases, dual-specificity kinases, plant kinases, bacterial kinases) in a single rule. The mixture of broad InterPro domains and specific CATH FunFam families suggests inconsistent curation criteria and incremental rule growth without systematic design principles.'
literature_support:
  assessment: WEAK
  notes: 'Deep research reveals that while individual kinase subfamily identification is well-supported in the literature, combining diverse kinase types in a single rule contradicts established kinase classification principles. The literature emphasizes that receptor tyrosine kinases, non-receptor tyrosine kinases, dual-specificity kinases, and bacterial kinases have distinct evolutionary origins, domain architectures, and functional roles that require separate annotation approaches. The use of generic "tyrosine-protein kinase" keywords for proteins that may include serine/threonine kinases and dual-specificity kinases is biologically inaccurate. Current best practices mandate using specific GO molecular function terms rather than broad UniProt keywords for automated annotation.'
  supported_by:
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137-deep-research-manual.md
    supporting_text: 'The classification of protein kinases into accurate functional subfamilies presents significant computational challenges due to: Domain Architecture Complexity, Sequence Divergence, and Functional Diversity. The term "tyrosine kinase" encompasses diverse proteins including receptor tyrosine kinases (RTKs), non-receptor tyrosine kinases (NRTKs), and dual-specificity kinases that can phosphorylate both tyrosine and serine/threonine residues.'
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137.enriched.json
    supporting_text: 'Rule contains annotations: [{annotationType: ANNOTATION, keyword: {id: KW-0829, name: Tyrosine-protein kinase, category: Unknown}}] - provides only generic keyword annotation rather than specific functional terms'
condition_overlap:
  assessment: SIGNIFICANT
  notes: 'With 98 condition sets requiring 4,753 pairwise comparisons, systematic overlap analysis is computationally prohibited. However, inspection of the condition sets reveals concerning patterns: the rule uses 35 InterPro domains and 106 CATH FunFam families in various combinations, creating potential redundancy. Many CATH FunFam families in the 1.10.510.10 and 3.30.200.20 superfamilies likely capture similar kinase types through different evolutionary clustering. Deep research indicates that broad InterPro families like IPR000719 (Protein kinase domain) are extremely promiscuous and likely captured by multiple condition sets with different additional constraints, suggesting systematic redundancy in protein targeting.'
  supported_by:
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137-deep-research-manual.md
    supporting_text: 'Many InterPro families used in kinase annotation (e.g., IPR000719 "Protein kinase domain") are extremely broad and capture hundreds of distinct kinase subfamilies. While CATH functional families (FunFams) provide more specific clustering than InterPro families, they still require careful validation to ensure functional coherence.'
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137.enriched.json
    supporting_text: 'Rule contains 98 condition sets with 35 unique InterPro domains, 106 unique CATH FunFam domains, and 4 PANTHER domains - the large number makes systematic overlap analysis infeasible'
go_specificity:
  assessment: TOO_BROAD
  notes: 'The rule applies only a generic UniProt keyword (KW-0829: Tyrosine-protein kinase) rather than specific GO molecular function terms, which severely limits its utility for computational functional analysis. Deep research reveals this is problematic because: (1) the rule captures diverse kinase types including dual-specificity kinases, plant receptor-like kinases, and bacterial kinases that should not receive a "tyrosine-protein kinase" annotation; (2) appropriate GO molecular function terms exist (GO:0004713 protein tyrosine kinase activity, GO:0004674 protein serine/threonine kinase activity, GO:0004715 non-membrane spanning protein tyrosine kinase activity) that would provide standardized, computable functional annotations; (3) the broad keyword fails to distinguish between functionally distinct kinase subfamilies with different biological roles and mechanisms.'
  supported_by:
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137-deep-research-manual.md
    supporting_text: 'Replace keyword annotations with specific GO molecular function terms like: GO:0004713: protein tyrosine kinase activity, GO:0004674: protein serine/threonine kinase activity, GO:0004715: non-membrane spanning protein tyrosine kinase activity'
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137.enriched.json
    supporting_text: 'Rule annotation: keyword {id: KW-0829, name: Tyrosine-protein kinase} but condition sets include dual-specificity kinases, serine/threonine kinases, and other non-tyrosine kinases'
taxonomic_scope:
  assessment: TOO_BROAD
  notes: 'The rule applies across an inappropriately broad taxonomic range spanning 15 different taxonomic groups from bacteria (Actinomycetota) to specific mammalian lineages (Primates, Glires). Deep research reveals that only 17 of 98 condition sets have taxonomic restrictions, leading to systematic over-annotation. The evolutionary distribution of kinase subfamilies is highly lineage-specific: receptor tyrosine kinases are primarily metazoan innovations, plant receptor-like kinases have distinct architectures from animal kinases, and bacterial kinases typically have different signaling contexts. The rule inappropriately applies the same "tyrosine-protein kinase" keyword to bacterial kinases, plant kinases, and animal kinases despite their fundamentally different evolutionary origins and functional contexts.'
  supported_by:
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137-deep-research-manual.md
    supporting_text: 'Receptor Tyrosine Kinases: Primarily metazoan innovations, with some exceptions. Applying RTK annotations to bacteria or plants would be inappropriate. Bacterial Kinases: Most bacterial kinases are involved in signaling rather than metabolism and have different domain architectures from eukaryotic kinases. Plant-Specific Kinases: Plants have evolved unique kinase subfamilies (e.g., receptor-like kinases) that are functionally distinct from animal kinases.'
  - reference_id: file:rules/arba/ARBA00023137/ARBA00023137.enriched.json
    supporting_text: 'Rule uses 15 different taxa (Actinomycetota, Amoebozoa, Bacteria, Chordata, Ecdysozoa, Embryophyta, Eukaryota, Eutheria, Glires, Metazoa, Primates, Pterygota, Streptophyta, Tracheophyta, Viridiplantae) with inconsistent application across kinase subfamilies'
confidence: 0.95
references:
- id: file:rules/arba/ARBA00023137/ARBA00023137.enriched.json
  title: ARBA rule enriched data
  findings:
  - statement: 'Rule contains 98 condition sets with 35 InterPro domains, 106 CATH FunFam families, and 4 PANTHER families, exceeding maintainability thresholds'
  - statement: 'Rule applies only generic keyword annotation (KW-0829) rather than specific GO molecular function terms'
  - statement: 'Rule mixes diverse kinase types inappropriately under single "tyrosine-protein kinase" annotation'
  - statement: 'Taxonomic scope spans 15 groups from bacteria to mammals with only 17/98 condition sets having restrictions'
- id: file:rules/arba/ARBA00023137/ARBA00023137-deep-research-manual.md
  title: Deep research on kinase classification and annotation challenges
  findings:
  - statement: 'Kinase classification requires subfamily-specific approaches due to domain architecture complexity and functional diversity'
  - statement: 'Receptor tyrosine kinases, non-receptor kinases, and bacterial kinases have distinct evolutionary origins requiring separate annotation'
  - statement: 'Broad InterPro families like IPR000719 are extremely promiscuous and unsuitable for accurate kinase subtype annotation'
  - statement: 'Current best practices require specific GO molecular function terms rather than generic UniProt keywords'
supported_by:
- reference_id: file:rules/arba/ARBA00023137/ARBA00023137.enriched.json
  supporting_text: 'Rule statistics: reviewedProteinCount: 0, unreviewedProteinCount: 203241, with 98 condition sets containing diverse kinase domain signatures across 35 InterPro domains, 106 CATH FunFam families, and 4 PANTHER families'
- reference_id: file:rules/arba/ARBA00023137/ARBA00023137-deep-research-manual.md
  supporting_text: 'This analysis has high confidence (0.85) based on: Well-established principles of kinase classification in the literature, Clear problems with the current rule structure, Specific, actionable recommendations for improvement, Alignment with established GO annotation best practices'