ARBA00023239

id: ARBA00023239
description: >
  Highly complex ARBA rule with 894 condition sets for detecting proteins with lyase activity
  based on InterPro domains, PANTHER families, and CATH FunFams across diverse taxonomic groups.
  Rule only assigns keyword annotation "Lyase" (KW-0456), not GO terms. Rule exceeds analysis
  limits due to excessive complexity, making quantitative validation impossible.

status: COMPLETE
rule_type: ARBA

rule:
  rule_id: ARBA00023239
  condition_sets: []
  go_annotations: []
  reviewed_protein_count: 0
  unreviewed_protein_count: 2585853
  created_date: "2020-05-12"
  modified_date: "2025-05-15"
  entries: []

review_summary: >
  ARBA00023239 represents a fundamental failure in automated annotation design. With 894
  condition sets, this rule is incomprehensibly complex, unmaintainable, and causes significant
  performance issues. The rule attempts to capture lyase activity across all domains of life
  but provides only a generic "Lyase" keyword that lacks functional specificity. The excessive
  complexity (far exceeding the 12 condition set analysis limit) makes validation impossible,
  while the >2.5 million protein coverage with zero reviewed proteins indicates massive
  over-annotation. This rule exemplifies the problems with broad, kitchen-sink approaches
  to annotation and should be completely removed in favor of focused, mechanistically-coherent
  rules for specific lyase subfamilies.

action: DEPRECATE

action_rationale: >
  This rule exhibits fatal design flaws that render it unsuitable for continued use:

  1. EXCESSIVE COMPLEXITY: 894 condition sets exceed all practical limits for validation,
     maintenance, and performance. The analysis system caps at 12 condition sets - this
     rule is 75x that limit.

  2. GENERIC ANNOTATION: Assigns only "Lyase" keyword without functional specificity.
     Lyases (EC 4.x.x.x) include carboxy-lyases, hydro-lyases, ammonia-lyases, etc.
     with vastly different mechanisms and substrates.

  3. UNMAINTAINABLE: No human curator can review literature support for 894 different
     condition combinations across all taxonomic domains.

  4. PERFORMANCE IMPACT: Mentioned in geneontology/go-annotation#6035 as causing pipeline
     performance issues due to computational overhead.

  5. ZERO SPECIFICITY: >2.5 million proteins with zero reviewed entries indicates this
     rule captures everything remotely lyase-related without discrimination.

  The rule should be deprecated and replaced with focused rules for specific lyase families
  using appropriate GO molecular function terms.

suggested_modifications:
- Deprecate entire rule due to fundamental design flaws
- Replace with separate targeted rules for major lyase classes (EC 4.1-4.6)
- Use specific GO molecular function terms (e.g., GO:0016831 carboxy-lyase activity)
- Limit replacement rules to <20 condition sets each
- Focus on mechanistically coherent enzyme families rather than broad taxonomic coverage
- Include comprehensive literature validation for each focused rule

parsimony:
  assessment: OVERLY_COMPLEX
  notes: >
    894 condition sets represent the antithesis of parsimony. This is systematic
    over-engineering that makes the rule impossible to validate, maintain, or optimize.
    The number of conditions exceeds practical analysis limits by 75-fold, indicating
    fundamental design failure rather than thoroughness.

literature_support:
  assessment: WEAK
  notes: >
    Literature support cannot be properly evaluated due to the rule's excessive complexity.
    No comprehensive literature review is feasible for 894 different domain combinations
    across all taxonomic domains. The generic "Lyase" annotation provides no mechanistic
    specificity that could be validated against biochemical literature.

condition_overlap:
  assessment: SIGNIFICANT
  notes: >
    With 894 condition sets targeting lyase-related domains across overlapping taxonomic
    groups, extensive redundancy is mathematically certain. The analysis system cannot
    even process this level of complexity, but basic inspection reveals many condition
    sets combine identical domains with different taxonomic restrictions, creating
    obvious redundancies.

go_specificity:
  assessment: TOO_BROAD
  notes: >
    The rule does not assign GO terms, only a generic "Lyase" keyword. This provides
    minimal functional insight and lumps together mechanistically diverse enzymes.
    Appropriate annotation would use specific GO molecular function terms for distinct
    lyase mechanisms (carboxy-lyase, hydro-lyase, etc.).

taxonomic_scope:
  assessment: TOO_BROAD
  notes: >
    The rule spans all domains of life with taxonomic restrictions scattered across
    894 condition sets. This approach treats lyase activity as a single universal
    function rather than recognizing that different lyase families have distinct
    evolutionary origins and taxonomic distributions. The broad scope contributes
    to the rule's unwieldy complexity.

confidence: 0.98

references:
- id: file:rules/arba/ARBA00023239/ARBA00023239.json
  title: Raw rule data
  findings:
  - statement: Rule contains 894 condition sets combining InterPro domains, PANTHER families, FunFams, and taxonomic restrictions
  - statement: Rule assigns only keyword annotation "Lyase" (KW-0456) without GO terms
  - statement: Rule covers 2,585,853 unreviewed proteins with 0 reviewed proteins
- id: geneontology/go-annotation#6035
  title: GitHub issue mentioning performance problems with complex ARBA rules
  findings:
  - statement: Complex ARBA rules like this one cause significant performance issues in annotation pipelines

supported_by:
- reference_id: file:rules/arba/ARBA00023239/ARBA00023239.json
  supporting_text: "Rule has 894 condition sets, which exceeds the maximum of 12. Analysis is skipped for rules with too many condition sets"