Rare Disease Diagnosis Advisor

Systematic diagnosis support for rare diseases using phenotype matching, gene panel prioritization, and variant interpretation across Orphanet, OMIM, HPO, ClinVar, and structure-based analysis.

KEY PRINCIPLES:

1. Report-first approach - Create report file FIRST, update progressively
2. Phenotype-driven - Convert symptoms to HPO terms before searching
3. Multi-database triangulation - Cross-reference Orphanet, OMIM, OpenTargets
4. Evidence grading - Grade diagnoses by supporting evidence strength
5. Actionable output - Prioritized differential diagnosis with next steps
6. Genetic counseling aware - Consider inheritance patterns and family history
7. English-first queries - Always use English terms in tool calls (phenotype descriptions, gene names, disease names), even if the user writes in another language. Only try original-language terms as a fallback. Respond in the user's language

When to Use

Apply when user asks:

• "Patient has [symptoms], what rare disease could this be?"
• "Unexplained developmental delay with [features]"
• "WES found VUS in [gene], is this pathogenic?"
• "What genes should we test for [phenotype]?"
• "Differential diagnosis for [rare symptom combination]"

Report-First Approach (MANDATORY)

1. Create the report file FIRST: [PATIENT_ID]_rare_disease_report.md with all section headers and [Researching...] placeholders
2. Progressively update as you gather data
3. Output separate data files:
   • [PATIENT_ID]_gene_panel.csv - Prioritized genes for testing
   • [PATIENT_ID]_variant_interpretation.csv - If variants provided

Every finding MUST include source citation (ORPHA code, OMIM number, tool name).

See REPORT_TEMPLATE.md for the full template and example outputs for each phase.

Tool Parameter Corrections

Workflow Overview

Phase 1: Phenotype Standardization
  Convert symptoms to HPO terms, identify core vs. variable features, note onset/inheritance
      |
Phase 2: Disease Matching
  Search Orphanet, cross-reference OMIM, query DisGeNET -> Ranked differential diagnosis
      |
Phase 3: Gene Panel Identification
  Extract genes from top diseases, validate with ClinGen, check expression (GTEx)
      |
Phase 3.5: Expression & Tissue Context
  CELLxGENE cell-type expression, ChIPAtlas regulatory context
      |
Phase 3.6: Pathway Analysis
  KEGG pathways, Reactome processes, IntAct protein interactions
      |
Phase 4: Variant Interpretation (if provided)
  ClinVar lookup, gnomAD frequency, computational predictions (CADD, AlphaMissense, EVE, SpliceAI)
      |
Phase 5: Structure Analysis (for VUS)
  AlphaFold2 prediction, domain impact assessment (InterPro)
      |
Phase 6: Literature Evidence
  PubMed studies, BioRxiv/MedRxiv preprints, OpenAlex citation analysis
      |
Phase 7: Report Synthesis
  Prioritized differential, recommended testing, next steps

For detailed code examples and algorithms for each phase, see DIAGNOSTIC_WORKFLOW.md.

Phase Summaries

Phase 1: Phenotype Standardization

• Use HPO_search_terms(query=symptom) to convert each clinical description to HPO terms
• Classify features as Core (always present), Variable (>50%), Occasional (<50%), or Age-specific
• Record age of onset and family history for inheritance pattern hints

Phase 2: Disease Matching

• Orphanet: Orphanet_search_diseases(operation="search_diseases", query=keyword) then Orphanet_get_genes(operation="get_genes", orpha_code=code) for each hit
• OMIM: OMIM_search(operation="search", query=gene) then OMIM_get_entry and OMIM_get_clinical_synopsis for details
• DisGeNET: DisGeNET_search_gene(operation="search_gene", gene=symbol) for gene-disease association scores
• Score phenotype overlap: Excellent (>80%), Good (60-80%), Possible (40-60%), Unlikely (<40%)

Phase 3: Gene Panel Identification

• Extract genes from top candidate diseases
• ClinGen validation (critical): ClinGen_search_gene_validity, ClinGen_search_dosage_sensitivity, ClinGen_search_actionability
• ClinGen classification determines panel inclusion:
  • Definitive/Strong/Moderate: Include in panel
  • Limited: Include but flag
  • Disputed/Refuted: Exclude
• Expression: Use MyGene_query_genes for Ensembl ID, then GTEx_get_median_gene_expression to confirm tissue expression
• Prioritization scoring: Tier 1 (top disease gene +5), Tier 2 (multi-disease +3), Tier 3 (ClinGen Definitive +3), Tier 4 (tissue expression +2), Tier 5 (pLI >0.9 +1)

Phase 3.5: Expression & Tissue Context

• CELLxGENE: CELLxGENE_get_expression_data and CELLxGENE_get_cell_metadata for cell-type specific expression
• ChIPAtlas: ChIPAtlas_enrichment_analysis and ChIPAtlas_get_peak_data for regulatory context (TF binding)
• Confirms candidate genes are expressed in disease-relevant tissues/cells

Phase 3.6: Pathway Analysis

• KEGG: kegg_find_genes(query="hsa:{gene}") then kegg_get_gene_info for pathway membership
• IntAct: intact_search_interactions(query=gene, species="human") for protein-protein interactions
• Identify convergent pathways across candidate genes (strengthens candidacy)

Phase 4: Variant Interpretation (if provided)

• ClinVar: ClinVar_search_variants(query=hgvs) for existing classifications
• gnomAD: gnomAD_get_variant_frequencies(variant_id=id) for population frequency
  • Ultra-rare (<0.00001), Rare (<0.0001), Low frequency (<0.01), Common (likely benign)
• Computational predictions (for VUS):
  • CADD: CADD_get_variant_score - PHRED >=20 supports PP3
  • AlphaMissense: AlphaMissense_get_variant_score - pathogenic classification = strong PP3
  • EVE: EVE_get_variant_score - score >0.5 supports PP3
  • SpliceAI: SpliceAI_predict_splice - delta score >=0.5 indicates splice impact
• ACMG criteria: PVS1 (null variant), PS1 (same AA change), PM2 (absent from pop), PP3 (computational), BA1 (>5% AF)
• Consensus from 2+ concordant predictors strengthens PP3 evidence

Phase 5: Structure Analysis (for VUS)

• Perform when: VUS, missense in critical domain, novel variant, or additional evidence needed
• AlphaFold2: NvidiaNIM_alphafold2(sequence=seq, algorithm="mmseqs2") for structure prediction
• Domain impact: InterPro_get_protein_domains(accession=uniprot_id) to check functional domains
• Assess pLDDT confidence at variant position, domain location, structural role

Phase 6: Literature Evidence

• PubMed: PubMed_search_articles(query="disease AND genetics") for published studies
• Preprints: BioRxiv_search_preprints, ArXiv_search_papers(category="q-bio") for latest findings
• Citations: openalex_search_works for citation analysis of key papers
• Note: preprints are not peer-reviewed; flag accordingly

Phase 7: Report Synthesis

• Compile all phases into final report with evidence grading
• Provide prioritized differential diagnosis with next steps
• Include specialist referral suggestions and family screening recommendations

Evidence Grading

Completeness Checklist

Phase 1 (Phenotype): All symptoms as HPO terms, core vs. variable distinguished, onset documented, family history noted

Phase 2 (Disease Matching): >=5 candidates (or all matching), overlap % calculated, inheritance patterns, ORPHA + OMIM IDs

Phase 3 (Gene Panel): >=5 genes prioritized, ClinGen evidence level per gene, expression validated, testing strategy recommended

Phase 4 (Variants): ClinVar classification, gnomAD frequency, ACMG criteria applied, classification justified

Phase 5 (Structure): Structure predicted (if VUS), pLDDT reported, domain impact assessed, structural evidence summarized

Phase 6 (Recommendations): >=3 next steps, specialist referrals, family screening addressed

See CHECKLIST.md for the full interactive checklist.

Fallback Chains

Reference Files

• DIAGNOSTIC_WORKFLOW.md - Detailed code examples and algorithms for each phase
• REPORT_TEMPLATE.md - Report template, phase output examples, CSV formats
• TOOLS_REFERENCE.md - Complete tool documentation
• CHECKLIST.md - Interactive completeness checklist
• EXAMPLES.md - Worked diagnosis examples