Tooluniverse Clinical Trial Matching

Clinical Trial Matching for Precision Medicine

Transform patient molecular profiles and clinical characteristics into prioritized clinical trial recommendations. Searches ClinicalTrials.gov and cross-references with molecular databases (CIViC, OpenTargets, ChEMBL, FDA) to produce evidence-graded, scored trial matches.

KEY PRINCIPLES: 1. Report-first approach - Create report file FIRST, then populate progressively 2. Patient-centric - Every recommendation considers the individual patient's profile 3. Molecular-first matching - Prioritize trials targeting patient's specific biomarkers

Molecular Matching Priority

Match patients to trials by molecular profile FIRST (specific mutations), then by disease stage, then by prior treatments. A patient with EGFR L858R should match to EGFR-targeted trials regardless of other factors. 4. Evidence-graded - Every recommendation has an evidence tier (T1-T4) 5. Quantitative scoring - Trial Match Score (0-100) for every trial 6. Eligibility-aware - Parse and evaluate inclusion/exclusion criteria 7. Actionable output - Clear next steps, contact info, enrollment status 8. Source-referenced - Every statement cites the tool/database source 9. Completeness checklist - Mandatory section showing analysis coverage 10. English-first queries - Always use English terms in tool calls. Respond in user's language

LOOK UP, DON'T GUESS

When uncertain about any scientific fact, SEARCH databases first rather than reasoning from memory. A database-verified answer is always more reliable than a guess.

---

COMPUTE, DON'T DESCRIBE

When analysis requires computation (statistics, data processing, scoring, enrichment), write and run Python code via Bash. Don't describe what you would do — execute it and report actual results. Use ToolUniverse tools to retrieve data, then Python (pandas, scipy, statsmodels, matplotlib) to analyze it.

When to Use

Apply when user asks:

• "What clinical trials are available for my NSCLC with EGFR L858R?"
• "Patient has BRAF V600E melanoma, failed ipilimumab - what trials?"
• "Find basket trials for NTRK fusion"
• "Breast cancer with HER2 amplification, post-CDK4/6 inhibitor trials"
• "KRAS G12C colorectal cancer clinical trials"
• "Immunotherapy trials for TMB-high solid tumors"
• "Clinical trials near Boston for lung cancer"
• "What are my options after failing osimertinib for EGFR+ NSCLC?"

NOT for (use other skills instead):

• Single variant interpretation without trial focus -> Use tooluniverse-cancer-variant-interpretation
• Drug safety profiling -> Use tooluniverse-adverse-event-detection
• Target validation -> Use tooluniverse-drug-target-validation
• General disease research -> Use tooluniverse-disease-research

---

Input Parsing

Required Input

• Disease/cancer type: Free-text disease name (e.g., "non-small cell lung cancer", "melanoma")

Strongly Recommended

• Molecular alterations: One or more biomarkers (e.g., "EGFR L858R", "KRAS G12C", "PD-L1 50%", "TMB-high")
• Stage/grade: Disease stage (e.g., "Stage IV", "metastatic", "locally advanced")
• Prior treatments: Previous therapies and outcomes (e.g., "failed platinum chemotherapy", "progressed on osimertinib")

Optional

• Performance status: ECOG or Karnofsky score
• Geographic location: City/state for proximity filtering
• Trial phase preference: I, II, III, IV, or "any"
• Intervention type: drug, biological, device, etc.
• Recruiting status preference: recruiting, not yet recruiting, active

For biomarker parsing rules and gene symbol normalization, see MATCHING_ALGORITHMS.md.

---

Workflow Overview

Input: Patient profile (disease + biomarkers + stage + prior treatments)

Phase 1: Patient Profile Standardization
  - Resolve disease to EFO/ontology IDs (OpenTargets, OLS)
  - Parse molecular alterations to gene + variant
  - Resolve gene symbols to Ensembl/Entrez IDs (MyGene)
  - Classify biomarker actionability (FDA-approved vs investigational)

Phase 2: Broad Trial Discovery
  - Disease-based trial search (ClinicalTrials.gov)
  - Biomarker-specific trial search
  - Intervention-based search (for known drugs targeting patient's biomarkers)
  - Deduplicate and collect NCT IDs

Phase 3: Trial Characterization (batch, groups of 10)
  - Eligibility criteria, conditions/interventions, locations, status, descriptions

Phase 4: Molecular Eligibility Matching
  - Parse eligibility text for biomarker requirements
  - Match patient's molecular profile to trial requirements
  - Score molecular eligibility (0-40 points)

Phase 5: Drug-Biomarker Alignment
  - Identify trial intervention drugs and mechanisms (OpenTargets, ChEMBL)
  - FDA approval status for biomarker-drug combinations
  - Classify drugs (targeted therapy, immunotherapy, chemotherapy)

Phase 6: Evidence Assessment
  - FDA-approved biomarker-drug combinations
  - Clinical trial results (PubMed), CIViC evidence, PharmGKB
  - Evidence tier classification (T1-T4)

Phase 7: Geographic & Feasibility Analysis
  - Trial site locations, enrollment status, proximity scoring

Phase 8: Alternative Options
  - Basket trials, expanded access, related studies

Phase 9: Scoring & Ranking (0-100 composite score)
  - Tier classification: Optimal (80-100) / Good (60-79) / Possible (40-59) / Exploratory (0-39)

Phase 10: Report Synthesis
  - Executive summary, ranked trial list, evidence grading, completeness checklist

---

Critical Tool Parameters

Clinical Trial Search Tools

Geographic coverage: ClinicalTrials.gov is U.S.-centric but global; many EU/EEA-only trials appear only in the EU CTIS register, and UK/international trials in ISRCTN. For a comprehensive search — or any patient who could enroll outside the U.S. — query search_clinical_trials, CTIS_search_trials, and ISRCTN_search_trials, then merge (the three registers list largely disjoint trials; ISRCTN records carry DOI/EudraCT/NCT cross-refs you can use to dedupe against the others). Each register has its own id namespace and detail tool: NCT→get_clinical_trial_*, CT number→CTIS_get_trial, ISRCTN id→ISRCTN_get_trial.

Batch Trial Detail Tools (all take nct_ids array)

Gene/Disease Resolution

Drug Information

Evidence Tools

Known CIViC Gene IDs

EGFR=19, BRAF=5, ALK=1, ABL1=4, KRAS=30, TP53=45, ERBB2=20, NTRK1=197, NTRK2=560, NTRK3=561, PIK3CA=37, MET=52, ROS1=118, RET=122, BRCA1=2370, BRCA2=2371

Critical Parameter Notes

1. DrugBank tools: ALL 4 parameters (query, case_sensitive, exact_match, limit) are REQUIRED 2. `search_clinical_trials`: query_term is REQUIRED even for disease-only searches 3. `search_clinical_trials`: action must be exactly "search_studies" 4. CIViC `civic_search_variants`: Does NOT filter by query - returns alphabetically 5. CIViC `civic_get_variants_by_gene`: Takes CIViC gene ID (integer), NOT gene symbol 6. Batch clinical trial tools: Accept arrays of NCT IDs, process in batches of 10

---

Scoring Summary

Trial Match Score (0-100):

• Molecular Match: 0-40 pts (exact variant=40, gene-level=30, pathway=20, none=10, excluded=0)
• Clinical Eligibility: 0-25 pts (all met=25, most=18, some=10, ineligible=0)
• Evidence Strength: 0-20 pts (FDA-approved=20, Phase III=15, Phase II=10, Phase I=5)
• Trial Phase: 0-10 pts (III=10, II=8, I/II=6, I=4)
• Geographic: 0-5 pts (local=5, same country=3, international=1)

Recommendation Tiers: Optimal (80-100), Good (60-79), Possible (40-59), Exploratory (0-39)

Evidence Tiers: T1 (FDA/guideline), T2 (Phase III), T3 (Phase I/II), T4 (computational)

For detailed scoring logic, see SCORING_CRITERIA.md.

---

Parallelization Strategy

Group 1 (Phase 1 - simultaneous):

• MyGene_query_genes per gene, OpenTargets disease search, ols_search_efo_terms, fda_pharmacogenomic_biomarkers

Group 2 (Phase 2 - simultaneous):

• search_clinical_trials by disease, biomarker, and intervention; search_clinical_trials alternative

Group 3 (Phase 3 - simultaneous):

• All batch detail tools (eligibility, interventions, locations, status, descriptions)

Group 4 (Phases 5-6 - per drug):

• Drug resolution, MoA, FDA indications, PubMed evidence

---

Error Handling

1. Wrap every tool call in try/except 2. Check for empty results and string error responses 3. Use fallback tools when primary fails (e.g., OLS if OpenTargets fails) 4. Document failures in completeness checklist 5. Never let one failure block the entire analysis

---

Reference Files

Clinical Trial Matching - Examples

Example 1: NSCLC with EGFR L858R

User Query: "Find clinical trials for Stage IV non-small cell lung cancer with EGFR L858R mutation, failed platinum chemotherapy, ECOG 0-1"

Skill Execution:

Phase 1: Patient Profile Standardization

• Disease: "non-small cell lung cancer" -> EFO_0003060 (non-small cell lung carcinoma)
• Gene: EGFR -> Ensembl: ENSG00000146648, Entrez: 1956
• Biomarker: EGFR L858R -> type: mutation, gene: EGFR, alteration: L858R
• FDA Biomarker: EGFR is FDA-recognized biomarker (afatinib, osimertinib, erlotinib, gefitinib)

Phase 2: Trial Discovery

• Disease search: ~200+ trials for NSCLC
• Biomarker search: "EGFR L858R" -> 10-30 targeted trials
• Intervention search: "osimertinib" for NSCLC -> 20+ trials
• Combined and deduplicated: ~50-80 unique candidate trials

Phase 3-6: Analysis

• Parse eligibility criteria for EGFR L858R requirements
• Check drug mechanisms for EGFR inhibitors
• Cross-reference FDA approvals (osimertinib for EGFR L858R NSCLC)
• Score molecular match: 40/40 for trials requiring EGFR L858R

Expected Top Matches

Trials testing:

• EGFR TKIs (osimertinib, amivantamab-vmjw)
• EGFR TKI combinations (TKI + chemotherapy)
• Next-generation EGFR inhibitors
• EGFR-targeted ADCs or bispecific antibodies

---

Example 2: Melanoma with BRAF V600E

User Query: "Patient has BRAF V600E melanoma, failed ipilimumab, looking for trials"

Skill Execution:

Phase 1: Standardization

• Disease: melanoma -> EFO_0000756
• Gene: BRAF -> ENSG00000157764, CIViC ID: 5
• Biomarker: BRAF V600E -> type: mutation
• FDA: BRAF is FDA-recognized (dabrafenib, vemurafenib, encorafenib)

Phase 2: Trial Discovery

• "melanoma BRAF V600E" -> BRAF-targeted trials
• "melanoma immunotherapy" -> post-ipilimumab options
• Combined targeted + immunotherapy options

Expected Top Matches

• BRAF/MEK inhibitor combinations (dabrafenib + trametinib)
• Novel BRAF inhibitors
• Combination of BRAF inhibition + immunotherapy
• Anti-PD-1 after ipilimumab
• Novel immunotherapy combinations

---

Example 3: Colorectal Cancer with KRAS G12C

User Query: "KRAS G12C colorectal cancer, Stage IV, failed FOLFOX + bevacizumab, microsatellite stable"

Skill Execution:

Phase 1: Standardization

• Disease: colorectal cancer -> EFO_0005842
• Gene: KRAS -> ENSG00000133703, CIViC ID: 30
• Biomarkers: KRAS G12C (mutation), MSS (status)
• FDA: KRAS G12C targeted (sotorasib approved for NSCLC, under investigation for CRC)

Phase 2: Trial Discovery

• "colorectal cancer KRAS G12C" -> specific trials
• "KRAS G12C" without disease filter -> cross-tumor trials
• "sotorasib" or "adagrasib" for CRC

Expected Top Matches

• KRAS G12C inhibitor trials (sotorasib, adagrasib) for CRC
• KRAS G12C + anti-EGFR combinations
• Novel KRAS inhibitors
• Basket trials for KRAS G12C any solid tumor

---

Example 4: Breast Cancer with HER2+

User Query: "HER2-positive breast cancer, failed trastuzumab + pertuzumab, brain metastases"

Skill Execution:

Phase 1: Standardization

• Disease: breast cancer -> EFO_0000305
• Gene: HER2/ERBB2 -> ENSG00000141736, CIViC ID: 20
• Biomarker: HER2 amplification
• FDA: HER2 recognized (trastuzumab, pertuzumab, T-DXd, tucatinib)

Phase 2: Trial Discovery

• "breast cancer HER2" + "brain metastases"
• "trastuzumab deruxtecan" (T-DXd) trials
• "tucatinib" trials (CNS-penetrant)
• Post-pertuzumab HER2+ trials

Special Considerations

• Brain metastases eligibility (many trials exclude CNS disease)
• CNS-active agents prioritized (tucatinib, T-DXd)

---

Example 5: NTRK Fusion (Basket Trial)

User Query: "Find trials for NTRK1 fusion, any solid tumor type"

Skill Execution:

Phase 1: Standardization

• Gene: NTRK1 -> ENSG00000198400, CIViC ID: 197
• Biomarker: NTRK1 fusion
• FDA: NTRK recognized (larotrectinib, entrectinib - tissue-agnostic)

Phase 2: Trial Discovery

• "NTRK fusion" + "basket" or "tumor agnostic"
• "larotrectinib" and "entrectinib" trials
• "NTRK" across all cancer types

Expected Top Matches

• Larotrectinib/entrectinib expansion studies
• Next-generation TRK inhibitors (selitrectinib, repotrectinib)
• Basket trials requiring NTRK fusion
• Combination approaches with TRK inhibition

---

Example 6: Geographic Search

User Query: "Find Phase II or III lung cancer trials near Boston"

Skill Execution:

Phase 1: Standardization

• Disease: lung cancer -> EFO_0001071
• Geographic preference: Boston, Massachusetts

Phase 2: Trial Discovery

• "lung cancer" Phase II/III
• Filter by Massachusetts locations

Phase 3: Location Analysis

• Sites in Massachusetts: Dana-Farber, Massachusetts General, Beth Israel
• Nearby states: Connecticut, Rhode Island, New Hampshire

Scoring

• Geographic feasibility weighted more heavily
• Massachusetts sites: 5/5 points
• Nearby New England sites: 3/5 points

---

Output Format

Each example produces a comprehensive markdown report saved to a file with the following key sections:

1. Executive Summary with top 3 trials and scores 2. Patient Profile with standardized identifiers 3. Ranked Trial List with scoring breakdown 4. Trial Categories (targeted, immunotherapy, combination, basket) 5. Alternative Options (expanded access, off-label) 6. Evidence Grading and completeness checklist

Matching Algorithms

Detailed algorithms for patient profile standardization, biomarker parsing, molecular eligibility matching, and drug-biomarker alignment.

Phase 1: Patient Profile Standardization

1.1 Disease Resolution

def resolve_disease(tu, disease_name):
    """Resolve disease name to EFO ID and standard terminology."""
    # OpenTargets disease search
    result = tu.tools.OpenTargets_get_disease_id_description_by_name(diseaseName=disease_name)
    hits = result.get('data', {}).get('search', {}).get('hits', [])

    if hits:
        disease_info = hits[0]
        return {
            'efo_id': disease_info.get('id'),
            'name': disease_info.get('name'),
            'description': disease_info.get('description'),
            'original_input': disease_name
        }

    # Fallback: OLS EFO search
    ols_result = tu.tools.ols_search_efo_terms(query=disease_name, limit=5)
    ols_terms = ols_result.get('data', {}).get('terms', [])
    if ols_terms:
        term = ols_terms[0]
        return {
            'efo_id': term.get('short_form'),
            'name': term.get('label'),
            'description': term.get('description', [''])[0] if term.get('description') else '',
            'original_input': disease_name
        }

    return {'efo_id': None, 'name': disease_name, 'description': '', 'original_input': disease_name}

1.2 Gene/Biomarker Resolution

def resolve_gene(tu, gene_symbol):
    """Resolve gene symbol to cross-database IDs."""
    alias_map = {
        'HER2': 'ERBB2', 'HER-2': 'ERBB2',
        'PD-L1': 'CD274', 'PDL1': 'CD274',
        'PD-1': 'PDCD1', 'PD1': 'PDCD1',
        'VEGF': 'VEGFA',
    }
    normalized = alias_map.get(gene_symbol.upper(), gene_symbol)

    result = tu.tools.MyGene_query_genes(query=normalized, species='human')
    hits = result.get('hits', [])

    gene_hit = None
    for hit in hits:
        if hit.get('symbol', '').upper() == normalized.upper():
            gene_hit = hit
            break
    if not gene_hit and hits:
        gene_hit = hits[0]

    if gene_hit:
        ensembl = gene_hit.get('ensembl', {})
        ensembl_id = ensembl.get('gene') if isinstance(ensembl, dict) else (ensembl[0].get('gene') if isinstance(ensembl, list) and ensembl else None)
        return {
            'symbol': gene_hit.get('symbol'),
            'entrez_id': gene_hit.get('entrezgene'),
            'ensembl_id': ensembl_id,
            'name': gene_hit.get('name'),
            'original_input': gene_symbol
        }

    return {'symbol': gene_symbol, 'entrez_id': None, 'ensembl_id': None, 'name': None, 'original_input': gene_symbol}

1.3 Biomarker Actionability Classification

def classify_biomarker_actionability(tu, gene_symbol, alteration):
    """Classify biomarker as FDA-approved, guideline, or investigational."""
    fda_result = tu.tools.fda_pharmacogenomic_biomarkers()
    fda_biomarkers = fda_result.get('results', [])

    fda_match = [b for b in fda_biomarkers if gene_symbol.upper() in str(b.get('Biomarker', '')).upper()]

    if fda_match:
        return {
            'level': 'FDA-approved',
            'drugs': [b.get('Drug') for b in fda_match],
            'labeling_sections': [b.get('LabelingSection') for b in fda_match]
        }

    return {'level': 'investigational', 'drugs': [], 'labeling_sections': []}

1.4 Biomarker Parsing

def parse_biomarker(biomarker_text):
    """Parse free-text biomarker into structured components."""
    import re

    # Pattern: "GENE VARIANT" (e.g., "EGFR L858R")
    mutation_match = re.match(r'(\w+)\s+([A-Z]\d+[A-Z])', biomarker_text, re.IGNORECASE)
    if mutation_match:
        return {'gene': mutation_match.group(1), 'alteration': mutation_match.group(2), 'type': 'mutation'}

    # Pattern: "GENE exon N deletion/insertion"
    exon_match = re.match(r'(\w+)\s+exon\s+(\d+)\s+(\w+)', biomarker_text, re.IGNORECASE)
    if exon_match:
        return {'gene': exon_match.group(1), 'alteration': f'exon {exon_match.group(2)} {exon_match.group(3)}', 'type': 'exon_alteration'}

    # Pattern: "GENE1-GENE2 fusion" or "GENE1/GENE2"
    fusion_match = re.match(r'(\w+)[-/](\w+)\s*(fusion)?', biomarker_text, re.IGNORECASE)
    if fusion_match:
        return {'gene': fusion_match.group(2), 'alteration': f'{fusion_match.group(1)}-{fusion_match.group(2)}', 'type': 'fusion', 'partner': fusion_match.group(1)}

    # Pattern: "GENE amplification"
    amp_match = re.match(r'(\w+)\s+amplification', biomarker_text, re.IGNORECASE)
    if amp_match:
        return {'gene': amp_match.group(1), 'alteration': 'amplification', 'type': 'amplification'}

    # Pattern: "PD-L1 XX%"
    expression_match = re.match(r'([\w-]+)\s+(\d+%|high|low|positive|negative)', biomarker_text, re.IGNORECASE)
    if expression_match:
        return {'gene': expression_match.group(1), 'alteration': expression_match.group(2), 'type': 'expression'}

    # Pattern: "MSI-high", "TMB-high"
    status_match = re.match(r'(MSI|TMB|dMMR|MMR)[-\s]*(high|low|stable|deficient|proficient)', biomarker_text, re.IGNORECASE)
    if status_match:
        return {'gene': status_match.group(1), 'alteration': status_match.group(2), 'type': 'status'}

    # Fallback
    return {'gene': biomarker_text.split()[0], 'alteration': ' '.join(biomarker_text.split()[1:]), 'type': 'unknown'}

Gene Symbol Normalization

Biomarker Parsing Rules

Phase 4: Molecular Eligibility Matching

Extract Biomarker Requirements from Eligibility Text

def extract_biomarker_requirements(eligibility_text):
    """Extract biomarker requirements from eligibility criteria text."""
    import re

    requirements = {
        'required_biomarkers': [],
        'excluded_biomarkers': [],
        'biomarker_agnostic': False
    }

    if not eligibility_text:
        return requirements

    text_upper = eligibility_text.upper()

    inclusion_section = eligibility_text.split('Exclusion Criteria')[0] if 'Exclusion Criteria' in eligibility_text else eligibility_text
    exclusion_section = eligibility_text.split('Exclusion Criteria')[1] if 'Exclusion Criteria' in eligibility_text else ''

    gene_patterns = [
        r'(?:EGFR|KRAS|BRAF|ALK|ROS1|RET|MET|NTRK|HER2|ERBB2|PIK3CA|BRCA|PD-?L1|MSI|TMB|dMMR)',
    ]

    for pattern in gene_patterns:
        for match in re.finditer(pattern, inclusion_section, re.IGNORECASE):
            gene = match.group(0).upper()
            context = inclusion_section[max(0, match.start()-100):match.end()+100]
            requirements['required_biomarkers'].append({
                'gene': gene,
                'context': context.strip()
            })

        for match in re.finditer(pattern, exclusion_section, re.IGNORECASE):
            gene = match.group(0).upper()
            context = exclusion_section[max(0, match.start()-100):match.end()+100]
            requirements['excluded_biomarkers'].append({
                'gene': gene,
                'context': context.strip()
            })

    basket_terms = ['tumor-agnostic', 'histology-independent', 'basket', 'any solid tumor', 'all comers', 'biomarker-selected']
    if any(term in text_upper.lower() for term in basket_terms):
        requirements['biomarker_agnostic'] = True

    return requirements

Phase 5: Drug-Biomarker Alignment

Get Drug Mechanism Info

def get_drug_mechanism_info(tu, drug_name):
    """Get drug mechanism, targets, and approval status."""
    result = tu.tools.OpenTargets_get_drug_id_description_by_name(drugName=drug_name)
    hits = result.get('data', {}).get('search', {}).get('hits', [])

    if not hits:
        return {'drug_name': drug_name, 'chembl_id': None, 'mechanisms': [], 'is_approved': False}

    drug_info = hits[0]
    chembl_id = drug_info.get('id')

    moa_result = tu.tools.OpenTargets_get_drug_mechanisms_of_action_by_chemblId(chemblId=chembl_id)
    moa_rows = moa_result.get('data', {}).get('drug', {}).get('mechanismsOfAction', {}).get('rows', [])

    mechanisms = []
    for row in moa_rows:
        targets = row.get('targets', [])
        mechanisms.append({
            'mechanism': row.get('mechanismOfAction'),
            'action_type': row.get('actionType'),
            'target_name': row.get('targetName'),
            'target_genes': [t.get('approvedSymbol') for t in targets]
        })

    return {
        'drug_name': drug_name,
        'chembl_id': chembl_id,
        'description': drug_info.get('description'),
        'mechanisms': mechanisms,
        'is_approved': 'approved' in drug_info.get('description', '').lower()
    }

FDA Approval Evidence

def check_fda_approval(tu, drug_name, disease_name):
    """Check FDA approval status and labeled indications."""
    result = tu.tools.FDA_get_indications_by_drug_name(drug_name=drug_name, limit=3)

    indications = result.get('results', [])
    for ind in indications:
        ind_text = str(ind.get('indications_and_usage', ''))
        if any(term.lower() in ind_text.lower() for term in disease_name.split()):
            return {
                'approved': True,
                'indication_text': ind_text[:500],
                'brand_name': ind.get('openfda.brand_name', []),
                'evidence_tier': 'T1'
            }

    return {'approved': False, 'indication_text': '', 'brand_name': [], 'evidence_tier': 'T3'}

Literature Evidence

def get_literature_evidence(tu, gene, alteration, drug_name, disease_name):
    """Search PubMed for evidence of drug efficacy for this biomarker."""
    query = f'{gene} {alteration} {drug_name} {disease_name} clinical trial'
    result = tu.tools.PubMed_search_articles(query=query, max_results=5)
    articles = result if isinstance(result, list) else result.get('articles', [])
    return articles

CIViC Evidence

def get_civic_evidence(tu, gene_symbol, civic_gene_id):
    """Get CIViC clinical evidence for gene variants."""
    if not civic_gene_id:
        return []
    result = tu.tools.civic_get_variants_by_gene(gene_id=civic_gene_id, limit=100)
    variants = result.get('data', {}).get('gene', {}).get('variants', {}).get('nodes', [])
    return variants

Clinical Trial Matching - Quick Start Guide

Overview

This skill matches patients to optimal clinical trials based on their molecular profile, disease characteristics, and eligibility criteria. It searches ClinicalTrials.gov and cross-references molecular databases to produce evidence-graded, scored trial recommendations.

Basic Usage

Simply describe your patient profile and ask for trial matches:

Find clinical trials for a patient with non-small cell lung cancer,
EGFR L858R mutation, Stage IV, who failed first-line osimertinib.

Use Case Examples

1. Targeted Therapy Matching

Patient profile:
- Disease: Non-small cell lung cancer (adenocarcinoma)
- Biomarker: EGFR L858R mutation
- Stage: Stage IV, metastatic
- Prior treatment: Failed platinum-based chemotherapy
- ECOG: 0-1

Find the best clinical trial matches.

2. Immunotherapy Selection

Match clinical trials for:
- Melanoma, advanced/metastatic
- TMB-high (>10 mut/Mb)
- PD-L1 positive (TPS >= 50%)
- Failed ipilimumab/nivolumab combination
- ECOG 0

3. Basket Trial Identification

Find basket/tumor-agnostic trials for a patient with:
- Solid tumor (colorectal cancer)
- NTRK1 fusion detected by NGS
- No prior TRK inhibitor therapy

4. Post-Progression Options

Clinical trial options for:
- HR+/HER2- breast cancer
- Failed CDK4/6 inhibitor (palbociclib) + letrozole
- ESR1 Y537S mutation detected
- Bone and liver metastases

5. Novel Biomarker Trials

Find trials for:
- Colorectal cancer, Stage IV
- KRAS G12C mutation
- Failed FOLFOX + bevacizumab
- MSS (microsatellite stable)

6. Geographic Search

Find lung cancer clinical trials:
- Non-small cell lung cancer, any molecular subtype
- Prefer trials in Boston, Massachusetts area
- Currently recruiting
- Phase II or III only

What You Get

For each patient, the skill produces:

1. Executive Summary - Top 3 trial recommendations with Trial Match Scores 2. Patient Profile - Standardized disease/biomarker information with EFO and gene IDs 3. Biomarker Actionability - FDA-approved vs investigational status 4. Ranked Trial List - Up to 10+ trials with detailed scoring breakdown:

• Molecular Match (0-40 points)
• Clinical Eligibility (0-25 points)
• Evidence Strength (0-20 points)
• Trial Phase (0-10 points)
• Geographic Feasibility (0-5 points)

5. Trial Details - NCT ID, phase, status, interventions, eligibility, locations 6. Drug-Biomarker Alignment - Whether trial drugs target the patient's biomarkers 7. Evidence Grading - T1 (FDA-approved) through T4 (computational) 8. Alternative Options - Basket trials, expanded access, off-label options 9. Additional Testing - Biomarker tests that would unlock more trials 10. Completeness Checklist - What analyses were performed

Trial Match Score Guide

Tips for Best Results

1. Be specific about biomarkers - Include variant-level detail (e.g., "EGFR L858R" not just "EGFR mutation") 2. Include prior treatments - Post-progression trials need to know what failed 3. Specify stage - Many trials require specific disease stages 4. Add geographic preference - If location matters, include city/state 5. Mention performance status - ECOG score helps filter eligibility 6. List multiple biomarkers - Complex profiles help find the best-matched trials

Data Sources

Limitations

• Trial availability changes frequently; always verify current status at ClinicalTrials.gov
• Eligibility assessment is approximate; final determination is by trial investigators
• Geographic distance calculations are approximate (state/city level, not exact)
• Report is for informational/research purposes only; discuss with healthcare team
• Some trials may have enrollment caps not reflected in public data

Report Template

Use the following markdown structure for the final clinical trial matching report.

File Naming Convention

clinical_trial_matching_[DISEASE]_[BIOMARKER]_[DATE].md

Example: clinical_trial_matching_NSCLC_EGFR_L858R_2026-02-15.md

Template

# Clinical Trial Matching Report

**Patient**: [Disease type] with [biomarker(s)]
**Date**: [Current date]
**Trials Analyzed**: [N total] | **Top Matches**: [N with score >= 60]

---

## Executive Summary

**Top 3 Trial Recommendations**:

1. **[NCT ID]** - [Brief title] (Score: XX/100, Tier N)
   - Phase: [Phase], Status: [Status]
   - Why: [Key reason for match]

2. **[NCT ID]** - [Brief title] (Score: XX/100, Tier N)
   ...

3. **[NCT ID]** - [Brief title] (Score: XX/100, Tier N)
   ...

---

## Patient Profile Summary

| Parameter | Value | Standardized |
|-----------|-------|-------------|
| Disease | [input] | [EFO name] (EFO_XXXX) |
| Biomarker(s) | [input] | [gene: variant, type] |
| Stage | [input] | [standardized] |
| Prior Treatment | [input] | [standardized] |
| Performance Status | [input] | [ECOG score] |
| Location | [input] | [city, state] |

### Biomarker Actionability
| Biomarker | Actionability Level | FDA-Approved Drugs | Evidence |
|-----------|--------------------|--------------------|----------|
| [gene variant] | [FDA-approved/investigational] | [drugs] | [T1/T2/T3/T4] |

---

## Ranked Trial Matches

### Trial 1: [NCT ID] - [Title]

**Trial Match Score: XX/100** (Tier N: [Label])

| Component | Score | Details |
|-----------|-------|---------|
| Molecular Match | XX/40 | [explanation] |
| Clinical Eligibility | XX/25 | [explanation] |
| Evidence Strength | XX/20 | [explanation] |
| Trial Phase | XX/10 | [phase] |
| Geographic | XX/5 | [location info] |

**Trial Details**:
- **Phase**: [Phase]
- **Status**: [Recruiting/Active/etc.]
- **Sponsor**: [Sponsor]
- **Start Date**: [Date]
- **Estimated Completion**: [Date]

**Interventions**:
- [Drug name]: [Mechanism] | [Dosing info if available]
- [Comparator]: [Description]

**Molecular Eligibility Match**:
- Required biomarkers: [list]
- Patient match: [Exact/Gene-level/Pathway/None]
- Notes: [details]

**Clinical Eligibility Assessment**:
- Disease type: [Match/Mismatch]
- Stage: [Match/Mismatch/Unclear]
- Prior treatment: [Match/Mismatch/Unclear]
- Performance status: [Match/Mismatch/Unclear]

**Evidence for Efficacy**:
- FDA approval: [Yes/No for this indication]
- Clinical results: [Phase III/II/I data if available]
- Mechanism alignment: [Drug targets patient's biomarker: Yes/No]
- Literature: [Key references]

**Trial Sites** (first 5):
- [City, State, Country]
- ...

**Next Steps**: [Contact info, enrollment instructions]

[Repeat for each matched trial]

---

## Trials by Category

### Targeted Therapy Trials
[List trials with targeted agents matching patient's biomarkers]

### Immunotherapy Trials
[List immunotherapy trials, noting PD-L1/TMB/MSI requirements]

### Combination Therapy Trials
[List trials with drug combinations]

### Basket/Platform Trials
[List biomarker-agnostic or multi-arm trials]

---

## Additional Testing Recommendations

If the patient has not been tested for certain biomarkers, these trials would become relevant:

| Biomarker | Test Needed | Trials Unlocked | Priority |
|-----------|-------------|----------------|----------|
| [e.g., TMB] | [NGS panel] | [NCT IDs] | [High/Medium/Low] |

---

## Alternative Options

### Expanded Access Programs
[List any expanded access or compassionate use programs]

### Off-Label Options
[FDA-approved drugs for other indications with same biomarker]

---

## Evidence Grading Summary

| Evidence Tier | Count | Description |
|--------------|-------|-------------|
| T1 (FDA/Guideline) | N | FDA-approved biomarker-drug, clinical guideline |
| T2 (Clinical) | N | Phase III data, robust clinical evidence |
| T3 (Emerging) | N | Phase I/II, preclinical evidence |
| T4 (Exploratory) | N | Computational, mechanism inference |

---

## Completeness Checklist

| Analysis Step | Status | Source |
|--------------|--------|--------|
| Disease standardization | [Done/Partial/Failed] | [OpenTargets/OLS] |
| Gene resolution | [Done/Partial/Failed] | [MyGene] |
| Biomarker actionability | [Done/Partial/Failed] | [FDA biomarkers] |
| Disease trial search | [Done/Partial/Failed] | [ClinicalTrials.gov] |
| Biomarker trial search | [Done/Partial/Failed] | [ClinicalTrials.gov] |
| Intervention trial search | [Done/Partial/Failed] | [ClinicalTrials.gov] |
| Eligibility parsing | [Done/Partial/Failed] | [ClinicalTrials.gov] |
| Drug mechanism analysis | [Done/Partial/Failed] | [OpenTargets/ChEMBL] |
| Evidence assessment | [Done/Partial/Failed] | [FDA/PubMed/CIViC] |
| Location analysis | [Done/Partial/Failed] | [ClinicalTrials.gov] |
| Basket trial search | [Done/Partial/Failed] | [ClinicalTrials.gov] |
| Expanded access search | [Done/Partial/Failed] | [ClinicalTrials.gov] |
| Scoring & ranking | [Done/Partial/Failed] | [Composite] |

---

## Disclaimer

This report is for informational and research purposes only. Clinical trial eligibility is ultimately determined by the trial investigators based on complete medical records. Patients should discuss all options with their healthcare team. Trial availability and status may change; verify current status at [ClinicalTrials.gov](https://clinicaltrials.gov).

## Sources

All data sourced from:
- ClinicalTrials.gov (trial search, eligibility, locations, status)
- OpenTargets Platform (drug-target associations, disease ontology)
- CIViC (clinical variant interpretations)
- ChEMBL (drug mechanisms, targets)
- FDA (approved indications, pharmacogenomic biomarkers, drug labels)
- DrugBank (drug targets, indications)
- PharmGKB (pharmacogenomics)
- PubMed/NCBI (literature evidence)
- OLS/EFO (disease ontology)
- MyGene (gene identifier resolution)

Scoring Criteria & Recommendation Tiers

Trial Match Score Components (Total: 0-100)

Molecular Match (0-40 points)

Clinical Eligibility (0-25 points)

Evidence Strength (0-20 points)

Trial Phase (0-10 points)

Geographic Feasibility (0-5 points)

Evidence Tier Classification

Recommendation Tiers

Molecular Match Scoring Logic

def score_molecular_match(patient_biomarkers, trial_requirements):
    """Score molecular match between patient and trial (0-40 points)."""
    if not trial_requirements['required_biomarkers'] and not trial_requirements['excluded_biomarkers']:
        return 10, 'No specific molecular criteria (general trial)'

    patient_genes = {b['gene'].upper() for b in patient_biomarkers}
    required_genes = {b['gene'].upper() for b in trial_requirements['required_biomarkers']}
    excluded_genes = {b['gene'].upper() for b in trial_requirements['excluded_biomarkers']}

    # Check exclusions first
    excluded_match = patient_genes & excluded_genes
    if excluded_match:
        return 0, f'Patient biomarker(s) {excluded_match} are in exclusion criteria'

    if not required_genes:
        return 10, 'No specific biomarker requirements found'

    # Check for exact gene match
    matched_genes = patient_genes & required_genes
    if matched_genes:
        exact_variant_match = False
        for req in trial_requirements['required_biomarkers']:
            for pb in patient_biomarkers:
                if pb['gene'].upper() == req['gene'].upper():
                    alt = pb.get('alteration', '').upper()
                    if alt and alt in req.get('context', '').upper():
                        exact_variant_match = True
                        break

        if exact_variant_match:
            return 40, f'Exact biomarker match: {matched_genes} with specific variant'
        else:
            return 30, f'Gene-level match: {matched_genes} (specific variant match unclear)'

    return 5, 'No direct biomarker match found'

Drug-Biomarker Alignment Scoring

def score_drug_biomarker_alignment(patient_gene_symbols, drug_mechanisms):
    """Check if trial drug targets patient's biomarkers."""
    patient_genes_upper = {g.upper() for g in patient_gene_symbols}

    for mech in drug_mechanisms:
        target_genes = {g.upper() for g in mech.get('target_genes', [])}
        if patient_genes_upper & target_genes:
            return True, f"Drug targets {patient_genes_upper & target_genes} via {mech.get('mechanism')}"

    return False, "No direct target overlap with patient biomarkers"

Clinical Trial Matching - Tools Reference

Summary

This skill uses 40+ ToolUniverse tools across 10 categories to perform comprehensive clinical trial matching.

Tool Inventory

Clinical Trial Tools (13 total, 11 primary)

OpenTargets Tools (10 primary)

CIViC Tools (6 primary)

FDA Tools (5 primary)

Gene/Disease Resolution Tools (4 primary)

Drug Information Tools (4 primary)

Literature Tools (2 primary)

PharmGKB Tools (2 primary)

Critical Parameter Notes

1. DrugBank tools: ALL 4 parameters (query, case_sensitive, exact_match, limit) are REQUIRED 2. `search_clinical_trials`: query_term is REQUIRED even for disease-only searches 3. `ClinicalTrials_search_studies`: action must be exactly "search_studies" 4. `ClinicalTrials_get_study`: action must be exactly "get_study_details" 5. CIViC `civic_search_variants`: Does NOT filter by query - returns alphabetically 6. CIViC `civic_get_variants_by_gene`: Takes CIViC gene ID (integer), NOT gene symbol 7. OpenTargets drug lookup: Use drugName (NOT genericName) 8. Batch clinical trial tools: Accept arrays of NCT IDs 9. `fda_pharmacogenomic_biomarkers`: Takes no parameters

Response Structure Cheat Sheet

search_clinical_trials:
  {studies: [{NCT ID, brief_title, brief_summary, overall_status, condition: [], phase: []}], nextPageToken, total_count}

ClinicalTrials_search_studies:
  {total_count, studies: [{nctId, title, status, conditions: []}]}

get_clinical_trial_eligibility_criteria:
  [{NCT ID, eligibility_criteria: "Inclusion Criteria:\n...\nExclusion Criteria:\n..."}]

get_clinical_trial_locations:
  [{NCT ID, locations: [{facility, city, state, country}]}]

get_clinical_trial_conditions_and_interventions:
  [{NCT ID, condition: [], arm_groups: [{label, type, description, interventionNames}], interventions: [{type, name, description}]}]

get_clinical_trial_status_and_dates:
  [{NCT ID, overall_status, start_date, primary_completion_date, completion_date}]

OpenTargets_get_drug_mechanisms_of_action_by_chemblId:
  {data: {drug: {id, name, mechanismsOfAction: {rows: [{mechanismOfAction, actionType, targetName, targets: [{id, approvedSymbol}]}]}}}}

OpenTargets_get_associated_drugs_by_target_ensemblID:
  {data: {target: {id, approvedSymbol, knownDrugs: {count, rows: [{drug: {id, name, isApproved}, phase, mechanismOfAction, disease: {id, name}}]}}}}

fda_pharmacogenomic_biomarkers:
  {count, shown, results: [{Drug, TherapeuticArea, Biomarker, LabelingSection}]}

MyGene_query_genes:
  {hits: [{symbol, entrezgene, ensembl: {gene}, name}]}

PubMed_search_articles:
  [{pmid, title, abstract, authors, journal, pub_date}]

Trial Search Patterns & Execution Strategy

Trial Search Functions

Disease-Based Search

def search_trials_by_disease(tu, disease_name, status_filter=None, phase_filter=None, page_size=20):
    """Search ClinicalTrials.gov by disease/condition."""
    query_parts = []
    if status_filter:
        query_parts.append(f'AREA[OverallStatus]{status_filter}')
    if phase_filter:
        query_parts.append(phase_filter)

    query_term = ' AND '.join(query_parts) if query_parts else disease_name

    result = tu.tools.search_clinical_trials(
        condition=disease_name,
        query_term=query_term if query_parts else disease_name,
        pageSize=page_size
    )

    if isinstance(result, str):
        return []
    return result.get('studies', [])

Biomarker-Specific Search

def search_trials_by_biomarker(tu, gene_symbol, alteration, disease_name=None, page_size=15):
    """Search trials mentioning specific biomarkers."""
    biomarker_query = f'{gene_symbol} {alteration}' if alteration else gene_symbol

    result = tu.tools.search_clinical_trials(
        condition=disease_name if disease_name else '',
        query_term=biomarker_query,
        pageSize=page_size
    )

    if isinstance(result, str):
        return []
    return result.get('studies', [])

Intervention-Based Search

def search_trials_by_intervention(tu, drug_name, disease_name=None, page_size=10):
    """Search trials by intervention/drug name."""
    result = tu.tools.search_clinical_trials(
        condition=disease_name if disease_name else '',
        intervention=drug_name,
        query_term=drug_name,
        pageSize=page_size
    )

    if isinstance(result, str):
        return []
    return result.get('studies', [])

Alternative Search (ClinicalTrials_search_studies)

def search_trials_alternative(tu, condition, intervention=None, limit=10):
    """Alternative trial search with different API endpoint."""
    params = {
        'action': 'search_studies',
        'condition': condition,
        'limit': limit
    }
    if intervention:
        params['intervention'] = intervention

    result = tu.tools.ClinicalTrials_search_studies(**params)
    return result.get('studies', [])

Deduplication

def deduplicate_trials(trial_lists):
    """Merge and deduplicate trials from multiple searches."""
    seen_ncts = set()
    unique_trials = []

    for trials in trial_lists:
        for trial in trials:
            nct = trial.get('NCT ID') or trial.get('nctId', '')
            if nct and nct not in seen_ncts:
                seen_ncts.add(nct)
                unique_trials.append(trial)

    return unique_trials

Batch Trial Detail Retrieval (Phase 3)

All batch functions process NCT IDs in groups of 10:

def get_trial_eligibility(tu, nct_ids):
    all_criteria = []
    for i in range(0, len(nct_ids), 10):
        batch = nct_ids[i:i+10]
        result = tu.tools.get_clinical_trial_eligibility_criteria(nct_ids=batch, eligibility_criteria='all')
        if isinstance(result, list):
            all_criteria.extend(result)
    return all_criteria

def get_trial_interventions(tu, nct_ids):
    all_interventions = []
    for i in range(0, len(nct_ids), 10):
        batch = nct_ids[i:i+10]
        result = tu.tools.get_clinical_trial_conditions_and_interventions(nct_ids=batch, condition_and_intervention='all')
        if isinstance(result, list):
            all_interventions.extend(result)
    return all_interventions

def get_trial_locations(tu, nct_ids):
    all_locations = []
    for i in range(0, len(nct_ids), 10):
        batch = nct_ids[i:i+10]
        result = tu.tools.get_clinical_trial_locations(nct_ids=batch, location='all')
        if isinstance(result, list):
            all_locations.extend(result)
    return all_locations

def get_trial_status(tu, nct_ids):
    all_status = []
    for i in range(0, len(nct_ids), 10):
        batch = nct_ids[i:i+10]
        result = tu.tools.get_clinical_trial_status_and_dates(nct_ids=batch, status_and_date='all')
        if isinstance(result, list):
            all_status.extend(result)
    return all_status

def get_trial_descriptions(tu, nct_ids):
    all_descriptions = []
    for i in range(0, len(nct_ids), 10):
        batch = nct_ids[i:i+10]
        result = tu.tools.get_clinical_trial_descriptions(nct_ids=batch, description_type='full')
        if isinstance(result, list):
            all_descriptions.extend(result)
    return all_descriptions

Geographic & Feasibility Analysis (Phase 7)

def analyze_trial_locations(locations_data, patient_location=None):
    """Analyze trial site locations and proximity."""
    if not locations_data:
        return {'total_sites': 0, 'countries': [], 'us_states': [], 'nearest': None}

    locations = locations_data.get('locations', [])
    countries = list(set(loc.get('country', '') for loc in locations if loc.get('country')))
    us_states = list(set(loc.get('state', '') for loc in locations if loc.get('country') == 'United States' and loc.get('state')))

    return {
        'total_sites': len(locations),
        'countries': countries,
        'us_states': us_states,
        'has_us_sites': 'United States' in countries,
        'locations': locations[:10]
    }

Alternative Options (Phase 8)

Basket Trial Search

IMPORTANT: ClinicalTrials.gov search is sensitive to query complexity. Overly specific queries like "NTRK fusion tumor agnostic" may return zero results. Use simpler queries and combine results.

def search_basket_trials(tu, biomarker, page_size=10):
    query_terms = [
        f'{biomarker} solid tumor',
        f'{biomarker}',
        f'{biomarker} basket',
    ]

    all_trials = []
    for query in query_terms:
        result = tu.tools.search_clinical_trials(query_term=query, pageSize=page_size)
        if not isinstance(result, str):
            all_trials.extend(result.get('studies', []))

    return deduplicate_trials([all_trials])

Expanded Access Search

def search_expanded_access(tu, drug_name):
    result = tu.tools.search_clinical_trials(query_term=f'{drug_name} expanded access', pageSize=5)
    if isinstance(result, str):
        return []
    return result.get('studies', [])

Parallelization Opportunities

Parallel Group 1 (Phase 1 - all simultaneous):

• MyGene_query_genes for each gene
• OpenTargets_get_disease_id_description_by_name for disease
• ols_search_efo_terms for disease
• fda_pharmacogenomic_biomarkers (no params)

Parallel Group 2 (Phase 2 - all simultaneous):

• search_clinical_trials with disease condition
• search_clinical_trials with biomarker query
• search_clinical_trials with intervention query
• ClinicalTrials_search_studies as alternative

Parallel Group 3 (Phase 3 - all simultaneous):

• get_clinical_trial_eligibility_criteria for all NCT IDs
• get_clinical_trial_conditions_and_interventions for all NCT IDs
• get_clinical_trial_locations for all NCT IDs
• get_clinical_trial_status_and_dates for all NCT IDs
• get_clinical_trial_descriptions for all NCT IDs

Parallel Group 4 (Phases 5-6 - for each drug):

• OpenTargets_get_drug_id_description_by_name for drug
• OpenTargets_get_drug_mechanisms_of_action_by_chemblId for drug
• FDA_get_indications_by_drug_name for drug
• PubMed_search_articles for evidence

Performance Optimization

• Batch NCT IDs in groups of 10 for detail tools
• Limit initial search to 20-30 trials per search strategy
• Focus detailed analysis on top 15-20 candidates after initial filtering
• Cache gene/disease resolution results for reuse across phases

Error Handling

For each tool call: 1. Wrap in try/except 2. Check for empty results 3. Use fallback tools when primary fails 4. Document what failed in completeness checklist 5. Never let one failure block the entire analysis

Common Use Patterns

Pattern 1: Targeted Therapy Matching (Most Common)

Input: "NSCLC patient with EGFR L858R, failed platinum chemotherapy" 1. Resolve: NSCLC -> EFO_0003060, EGFR -> ENSG00000146648 2. Search: "non-small cell lung cancer" + "EGFR mutation" + "EGFR L858R" 3. Filter: Recruiting trials with EGFR molecular requirements 4. Match: Score trials by EGFR L858R specificity 5. Drugs: Identify TKIs (osimertinib, erlotinib, etc.) in trial arms 6. Evidence: Check FDA approval of EGFR TKIs for NSCLC 7. Report: Prioritize targeted therapy trials, include immunotherapy options

Pattern 2: Immunotherapy Selection

Input: "Melanoma, TMB-high, PD-L1 positive, failed ipilimumab" 1. Resolve: Melanoma -> EFO_0000756 2. Search: "melanoma" + "TMB" + "PD-L1" + "immunotherapy" 3. Filter: Trials requiring PD-L1 or TMB testing 4. Match: Score by TMB/PD-L1 requirements 5. Drugs: Identify checkpoint inhibitors (pembrolizumab, nivolumab) 6. Evidence: Check FDA approval for TMB-high indications 7. Report: Focus on anti-PD-1/PD-L1 trials, combination immunotherapy

Pattern 3: Basket Trial Identification

Input: "Any solid tumor with NTRK fusion" 1. Resolve: NTRK genes (NTRK1, NTRK2, NTRK3) 2. Search: "NTRK fusion" + "tumor agnostic" + "basket" 3. Filter: Biomarker-agnostic trials 4. Match: Score by NTRK-specific inclusion criteria 5. Drugs: Identify larotrectinib, entrectinib 6. Evidence: FDA tissue-agnostic approval for larotrectinib 7. Report: Highlight tumor-agnostic approval, broad eligibility

Pattern 4: Post-Progression Options

Input: "Breast cancer, failed CDK4/6 inhibitors, ESR1 mutation" 1. Resolve: Breast cancer -> EFO_0000305, ESR1 -> ENSG00000091831 2. Search: "breast cancer" + "ESR1" + "CDK4/6 resistance" 3. Filter: Trials for post-CDK4/6 setting 4. Match: Score by ESR1 mutation and prior treatment requirements 5. Drugs: Identify novel endocrine agents, SERDs, ESR1-targeting drugs 6. Evidence: Check clinical data for post-CDK4/6 options 7. Report: Focus on resistance-overcoming strategies

Pattern 5: Geographic Search

Input: "Lung cancer trials within 100 miles of Boston" 1. Search: "lung cancer" (broad) 2. Get locations for all candidate trials 3. Filter: Sites in Massachusetts and nearby states 4. Score: High geographic feasibility for Boston-area sites 5. Report: Prioritize by proximity, include contact info

Edge Case Handling

No Matching Trials Found

1. Broaden search to gene-level (remove specific variant) 2. Search for pathway-level trials 3. Search basket trials 4. Suggest additional biomarker testing 5. Report alternative options (off-label, compassionate use)

Rare Biomarkers

1. Search gene-level trials (any EGFR mutation) 2. Search mechanism-level trials (TKI trials) 3. Check CIViC for any evidence on this specific variant 4. Note variant rarity in report 5. Suggest discussion with molecular tumor board

Multiple Biomarkers

1. Search for each biomarker independently 2. Search for combination biomarker trials 3. Identify trials that require multiple biomarkers 4. Score based on most actionable biomarker 5. Flag potential synergistic drug targets

Conflicting Eligibility

1. Score partial match transparently 2. Highlight which criteria are met/unmet 3. Note if unmet criteria are waivable 4. Suggest contacting PI for edge cases 5. Provide alternative trials without conflicting criteria