Skills

tooluniverse-adverse-event-detection

Originally frommims-harvard/tooluniverse
Installation
SKILL.md

Adverse Drug Event Signal Detection & Analysis

Automated pipeline for detecting, quantifying, and contextualizing adverse drug event signals using FAERS disproportionality analysis, FDA label mining, mechanism-based prediction, and literature evidence. Produces a quantitative Safety Signal Score (0-100) for regulatory and clinical decision-making.

KEY PRINCIPLES:

  1. Signal quantification first - Every adverse event must have PRR/ROR/IC with confidence intervals
  2. Serious events priority - Deaths, hospitalizations, life-threatening events always analyzed first
  3. Multi-source triangulation - FAERS + FDA labels + OpenTargets + DrugBank + literature
  4. Context-aware assessment - Distinguish drug-specific vs class-wide vs confounding signals
  5. Report-first approach - Create report file FIRST, update progressively
  6. Evidence grading mandatory - T1 (regulatory/boxed warning) through T4 (computational)
  7. English-first queries - Always use English drug names in tool calls, respond in user's language

When to Use

Apply when user asks:

  • "What are the safety signals for [drug]?"
  • "Detect adverse events for [drug]"
  • "Is [drug] associated with [adverse event]?"
  • "What are the FAERS signals for [drug]?"
  • "Compare safety of [drug A] vs [drug B] for [adverse event]"
  • "What are the serious adverse events for [drug]?"
  • "Are there emerging safety signals for [drug]?"
  • "Post-market surveillance report for [drug]"
  • "Pharmacovigilance signal detection for [drug]"
  • "What is the disproportionality analysis for [drug] and [event]?"

Differentiation from tooluniverse-pharmacovigilance: This skill focuses specifically on signal detection and quantification using disproportionality analysis (PRR, ROR, IC) with statistical rigor, produces a quantitative Safety Signal Score (0-100), and performs comparative safety analysis across drug classes. The pharmacovigilance skill provides broader safety profiling without the same depth of signal detection metrics.

Workflow Overview

Phase 0: Input Parsing & Drug Disambiguation
  Parse drug name, resolve to ChEMBL ID, DrugBank ID
  Identify drug class, mechanism, and approved indications
    |
Phase 1: FAERS Adverse Event Profiling
  Top adverse events by frequency
  Seriousness and outcome distributions
  Demographics (age, sex, country)
    |
Phase 2: Disproportionality Analysis (Signal Detection)
  Calculate PRR, ROR, IC with 95% CI for each AE
  Apply signal detection criteria
  Classify signal strength (Strong/Moderate/Weak/None)
    |
Phase 3: FDA Label Safety Information
  Boxed warnings, contraindications
  Warnings and precautions, adverse reactions
  Drug interactions, special populations
    |
Phase 4: Mechanism-Based Adverse Event Context
  Target-based AE prediction (OpenTargets safety)
  Off-target effects, ADMET predictions
  Drug class effects comparison
    |
Phase 5: Comparative Safety Analysis
  Compare to drugs in same class
  Identify unique vs class-wide signals
  Head-to-head disproportionality comparison
    |
Phase 6: Drug-Drug Interactions & Risk Factors
  Known DDIs causing AEs
  Pharmacogenomic risk factors (PharmGKB)
  FDA PGx biomarkers
    |
Phase 7: Literature Evidence
  PubMed safety studies, case reports
  OpenAlex citation analysis
  Preprint emerging signals (EuropePMC)
    |
Phase 8: Risk Assessment & Safety Signal Score
  Calculate Safety Signal Score (0-100)
  Evidence grading (T1-T4) for each signal
  Clinical significance assessment
    |
Phase 9: Report Synthesis & Recommendations
  Monitoring recommendations
  Risk mitigation strategies
  Completeness checklist

Phase 0: Input Parsing & Drug Disambiguation

0.1 Resolve Drug Identity

# Step 1: Get ChEMBL ID from drug name
chembl_result = tu.tools.OpenTargets_get_drug_chembId_by_generic_name(drugName="atorvastatin")
# Response: {data: {search: {hits: [{id: "CHEMBL1487", name: "ATORVASTATIN", description: "..."}]}}}
chembl_id = chembl_result['data']['search']['hits'][0]['id']  # "CHEMBL1487"

# Step 2: Get drug mechanism of action
moa = tu.tools.OpenTargets_get_drug_mechanisms_of_action_by_chemblId(chemblId=chembl_id)
# Response: {data: {drug: {mechanismsOfAction: {rows: [{mechanismOfAction: "HMG-CoA reductase inhibitor", actionType: "INHIBITOR", targetName: "...", targets: [{id: "ENSG00000113161", approvedSymbol: "HMGCR"}]}]}}}}

# Step 3: Get blackbox warning status
blackbox = tu.tools.OpenTargets_get_drug_blackbox_status_by_chembl_ID(chemblId=chembl_id)
# Response: {data: {drug: {name: "ATORVASTATIN", hasBeenWithdrawn: false, blackBoxWarning: false}}}

# Step 4: Get DrugBank info (safety, toxicity)
drugbank = tu.tools.drugbank_get_safety_by_drug_name_or_drugbank_id(
    query="atorvastatin", case_sensitive=False, exact_match=False, limit=3
)
# Response: {results: [{drug_name: "Atorvastatin", drugbank_id: "DB01076", toxicity: "...", food_interactions: "..."}]}

# Step 5: Get DrugBank targets
targets = tu.tools.drugbank_get_targets_by_drug_name_or_drugbank_id(
    query="atorvastatin", case_sensitive=False, exact_match=False, limit=3
)
# Response: {results: [{drug_name: "...", targets: [{name: "HMG-CoA reductase", ...}]}]}

# Step 6: Get approved indications
indications = tu.tools.OpenTargets_get_drug_indications_by_chemblId(chemblId=chembl_id)
# Response: {data: {drug: {indications: {rows: [{disease: {name: "hypercholesterolemia"}, maxPhaseForIndication: 4}]}}}}

0.2 Output for Report

## 1. Drug Identification

| Property | Value |
|----------|-------|
| **Generic Name** | Atorvastatin |
| **ChEMBL ID** | CHEMBL1487 |
| **DrugBank ID** | DB01076 |
| **Drug Class** | HMG-CoA reductase inhibitor (Statin) |
| **Mechanism** | HMG-CoA reductase inhibitor (target: HMGCR) |
| **Primary Target** | HMGCR (ENSG00000113161) |
| **Black Box Warning** | No |
| **Withdrawn** | No |

*Source: OpenTargets, DrugBank*

Phase 1: FAERS Adverse Event Profiling

1.1 Query FAERS for Adverse Events

# Get top adverse event reactions (returns list of {term, count})
reactions = tu.tools.FAERS_count_reactions_by_drug_event(medicinalproduct="ATORVASTATIN")
# Response: [{term: "FATIGUE", count: 19171}, {term: "DIARRHOEA", count: 17127}, ...]

# Get seriousness classification
seriousness = tu.tools.FAERS_count_seriousness_by_drug_event(medicinalproduct="ATORVASTATIN")
# Response: [{term: "Serious", count: 242757}, {term: "Non-serious", count: 83504}]

# Get outcome distribution
outcomes = tu.tools.FAERS_count_outcomes_by_drug_event(medicinalproduct="ATORVASTATIN")
# Response: [{term: "Unknown", count: 162310}, {term: "Fatal", count: 22128}, ...]

# Get age distribution
age_dist = tu.tools.FAERS_count_patient_age_distribution(medicinalproduct="ATORVASTATIN")
# Response: [{term: "Elderly", count: 38510}, {term: "Adult", count: 24302}, ...]

# Get death-related events
deaths = tu.tools.FAERS_count_death_related_by_drug(medicinalproduct="ATORVASTATIN")
# Response: [{term: "alive", count: 113157}, {term: "death", count: 26909}]

# Get reporter country distribution
countries = tu.tools.FAERS_count_reportercountry_by_drug_event(medicinalproduct="ATORVASTATIN")
# Response: [{term: "US", count: 170963}, {term: "GB", count: 40079}, ...]

1.2 Get Serious Events Breakdown

# Filter serious events - all types
serious_all = tu.tools.FAERS_filter_serious_events(
    operation="filter_serious_events",
    drug_name="ATORVASTATIN",
    seriousness_type="all"
)
# Response: {status: "success", drug_name: "ATORVASTATIN", seriousness_type: "all",
#   total_serious_events: N, top_serious_reactions: [{reaction: "...", count: N}, ...]}

# Death-related serious events
serious_death = tu.tools.FAERS_filter_serious_events(
    operation="filter_serious_events",
    drug_name="ATORVASTATIN",
    seriousness_type="death"
)
# Response: {status: "success", total_serious_events: 18720,
#   top_serious_reactions: [{reaction: "DEATH", count: 7541}, {reaction: "MYOCARDIAL INFARCTION", count: 1286}, ...]}

# Hospitalization-related
serious_hosp = tu.tools.FAERS_filter_serious_events(
    operation="filter_serious_events",
    drug_name="ATORVASTATIN",
    seriousness_type="hospitalization"
)

# Life-threatening
serious_lt = tu.tools.FAERS_filter_serious_events(
    operation="filter_serious_events",
    drug_name="ATORVASTATIN",
    seriousness_type="life_threatening"
)

1.3 MedDRA Hierarchy Rollup

# Get MedDRA preferred term rollup (top 50)
meddra = tu.tools.FAERS_rollup_meddra_hierarchy(
    operation="rollup_meddra_hierarchy",
    drug_name="ATORVASTATIN"
)
# Response: {status: "success", drug_name: "ATORVASTATIN",
#   meddra_hierarchy: {PT_level: [{preferred_term: "FATIGUE", count: 13957}, ...]}}

1.4 Output for Report

## 2. FAERS Adverse Event Profile

### 2.1 Overview
- **Total reports**: 326,261 (Serious: 242,757 | Non-serious: 83,504)
- **Fatal outcomes**: 22,128
- **Primary reporter countries**: US (170,963), GB (40,079), CA (16,492)

### 2.2 Top 10 Adverse Events by Frequency

| Rank | Adverse Event | Reports | % of Total |
|------|---------------|---------|------------|
| 1 | Fatigue | 19,171 | 5.9% |
| 2 | Diarrhoea | 17,127 | 5.2% |
| 3 | Dyspnoea | 15,992 | 4.9% |
| ... | ... | ... | ... |

### 2.3 Outcome Distribution

| Outcome | Count | Percentage |
|---------|-------|------------|
| Unknown | 162,310 | 39.6% |
| Recovered/resolved | 94,737 | 23.1% |
| Not recovered | 77,721 | 18.9% |
| Recovering | 49,367 | 12.0% |
| Fatal | 22,128 | 5.4% |
| Recovered with sequelae | 4,930 | 1.2% |

### 2.4 Age Distribution

| Age Group | Reports | Percentage |
|-----------|---------|------------|
| Elderly | 38,510 | 61.3% |
| Adult | 24,302 | 38.7% |
| Other | 152 | <1% |

*Source: FAERS via FAERS_count_reactions_by_drug_event, FAERS_count_seriousness_by_drug_event*

Phase 2: Disproportionality Analysis (Signal Detection)

2.1 Calculate Signal Metrics

CRITICAL: This is the core of the skill. For each top adverse event (at least top 15-20), calculate PRR, ROR, and IC with 95% confidence intervals.

# For each significant adverse event, calculate disproportionality
top_events = ["Rhabdomyolysis", "Myalgia", "Hepatotoxicity", "Diabetes mellitus",
              "Acute kidney injury", "Myopathy", "Pancreatitis"]

for event in top_events:
    result = tu.tools.FAERS_calculate_disproportionality(
        operation="calculate_disproportionality",
        drug_name="ATORVASTATIN",
        adverse_event=event
    )
    # Response structure:
    # {
    #   status: "success",
    #   drug_name: "ATORVASTATIN",
    #   adverse_event: "Rhabdomyolysis",
    #   contingency_table: {
    #     a_drug_and_event: 2226,
    #     b_drug_no_event: 241655,
    #     c_no_drug_event: 37658,
    #     d_no_drug_no_event: 19725450
    #   },
    #   metrics: {
    #     ROR: {value: 4.825, ci_95_lower: 4.622, ci_95_upper: 5.037},
    #     PRR: {value: 4.79, ci_95_lower: 4.59, ci_95_upper: 4.998},
    #     IC: {value: 2.194, ci_95_lower: 2.136, ci_95_upper: 2.252}
    #   },
    #   signal_detection: {
    #     signal_detected: true,
    #     signal_strength: "Strong signal",
    #     criteria: "ROR lower CI > 1.0 and case count >= 3"
    #   }
    # }

2.2 Signal Detection Criteria

Proportional Reporting Ratio (PRR):

  • PRR = (a/(a+b)) / (c/(c+d))
  • Signal: PRR >= 2.0 AND lower 95% CI > 1.0 AND case count >= 3

Reporting Odds Ratio (ROR):

  • ROR = (ad) / (bc)
  • Signal: Lower 95% CI > 1.0

Information Component (IC):

  • IC = log2(observed/expected)
  • Signal: Lower 95% CI > 0

2.3 Signal Strength Classification

Strength PRR ROR Lower CI IC Lower CI Clinical Action
Strong >= 5.0 >= 3.0 >= 2.0 Immediate investigation required
Moderate 3.0-4.9 2.0-2.9 1.0-1.9 Active monitoring recommended
Weak 2.0-2.9 1.0-1.9 0-0.9 Routine monitoring, watch for trends
No signal < 2.0 < 1.0 < 0 Standard pharmacovigilance

2.4 Demographic Stratification of Key Signals

# For strong/moderate signals, stratify by demographics
result = tu.tools.FAERS_stratify_by_demographics(
    operation="stratify_by_demographics",
    drug_name="ATORVASTATIN",
    adverse_event="Rhabdomyolysis",
    stratify_by="sex"  # Options: sex, age, country
)
# Response: {status: "success", total_reports: 1996,
#   stratification: [{group: 1, count: 1190, percentage: 59.62},  # 1=Male
#                    {group: 2, count: 781, percentage: 39.13}]}    # 2=Female

Note on sex codes: group 0 = Unknown, group 1 = Male, group 2 = Female.

2.5 Output for Report

## 3. Disproportionality Analysis (Signal Detection)

### 3.1 Signal Detection Summary

| Adverse Event | Cases (a) | PRR | PRR 95% CI | ROR | ROR 95% CI | IC | Signal |
|---------------|-----------|-----|------------|-----|------------|-----|--------|
| Rhabdomyolysis | 2,226 | 4.79 | 4.59-5.00 | 4.83 | 4.62-5.04 | 2.19 | **STRONG** |
| Myopathy | 1,234 | 6.12 | 5.72-6.55 | 6.18 | 5.77-6.62 | 2.54 | **STRONG** |
| Myalgia | 9,189 | 2.31 | 2.26-2.37 | 2.33 | 2.28-2.39 | 1.18 | Moderate |
| Hepatotoxicity | 456 | 3.45 | 3.10-3.84 | 3.48 | 3.13-3.87 | 1.72 | Moderate |
| Diabetes mellitus | 3,021 | 1.89 | 1.82-1.96 | 1.90 | 1.83-1.97 | 0.91 | Weak |
| Pancreatitis | 678 | 2.15 | 1.97-2.34 | 2.16 | 1.98-2.35 | 1.08 | Weak |

### 3.2 Demographics of Key Signals

**Rhabdomyolysis** (n=1,996):
- Male: 59.6%, Female: 39.1%, Unknown: 1.3%
- Predominantly elderly (>65 years)

*Source: FAERS via FAERS_calculate_disproportionality, FAERS_stratify_by_demographics*

Phase 3: FDA Label Safety Information

3.1 Extract Label Sections

# Boxed warnings
boxed = tu.tools.FDA_get_boxed_warning_info_by_drug_name(drug_name="atorvastatin")
# Response: {meta: {total: N}, results: [{boxed_warning: ["...text..."]}]}
# NOTE: Returns {error: {code: "NOT_FOUND"}} if no boxed warning exists

# Contraindications
contras = tu.tools.FDA_get_contraindications_by_drug_name(drug_name="atorvastatin")
# Response: {meta: {total: N}, results: [{openfda.generic_name: [...], contraindications: ["...text..."]}]}

# Warnings and precautions
warnings = tu.tools.FDA_get_warnings_by_drug_name(drug_name="atorvastatin")
# Response: {meta: {total: N}, results: [{warnings: ["...text..."], boxed_warning: [...]}]}

# Adverse reactions from label
adverse_rxns = tu.tools.FDA_get_adverse_reactions_by_drug_name(drug_name="atorvastatin")
# Response: {meta: {total: N}, results: [{adverse_reactions: ["...text..."]}]}

# Drug interactions from label
interactions = tu.tools.FDA_get_drug_interactions_by_drug_name(drug_name="atorvastatin")
# Response: {meta: {total: N}, results: [{drug_interactions: ["...text..."]}]}

# Pregnancy/breastfeeding
pregnancy = tu.tools.FDA_get_pregnancy_or_breastfeeding_info_by_drug_name(drug_name="atorvastatin")

# Geriatric use
geriatric = tu.tools.FDA_get_geriatric_use_info_by_drug_name(drug_name="atorvastatin")

# Pediatric use
pediatric = tu.tools.FDA_get_pediatric_use_info_by_drug_name(drug_name="atorvastatin")

# Pharmacogenomics from label
pgx_label = tu.tools.FDA_get_pharmacogenomics_info_by_drug_name(drug_name="atorvastatin")

3.2 Handling No Results

IMPORTANT: FDA label tools return {error: {code: "NOT_FOUND"}} when a section does not exist. This is NORMAL for many drugs - for example, most drugs do NOT have boxed warnings. Always check for this pattern:

# Check if boxed warning exists
if isinstance(boxed, dict) and 'error' in boxed:
    boxed_warning_text = "None (no boxed warning for this drug)"
else:
    boxed_warning_text = boxed['results'][0].get('boxed_warning', ['None'])[0]

3.3 Output for Report

## 4. FDA Label Safety Information

### 4.1 Boxed Warning
None

### 4.2 Contraindications
- Acute liver failure or decompensated cirrhosis
- Hypersensitivity to atorvastatin (includes anaphylaxis, angioedema, SJS, TEN)

### 4.3 Warnings and Precautions
| Warning | Clinical Relevance |
|---------|-------------------|
| Myopathy/Rhabdomyolysis | Risk with CYP3A4 inhibitors, high doses |
| Immune-Mediated Necrotizing Myopathy | Rare autoimmune myopathy |
| Hepatic Dysfunction | Monitor LFTs |
| Increased HbA1c/Glucose | Diabetes risk |

### 4.4 Drug Interactions (from label)
| Interacting Drug | Mechanism | Clinical Action |
|-----------------|-----------|-----------------|
| Cyclosporine | Increased exposure | Avoid combination |
| CYP3A4 inhibitors | Increased atorvastatin levels | Use lowest dose |
| Gemfibrozil | Increased myopathy risk | Avoid |

### 4.5 Special Populations
- **Pregnancy**: Contraindicated
- **Geriatric**: No dose adjustment needed
- **Pediatric**: Approved for heterozygous FH ages 10+

*Source: FDA drug labels via FDA_get_contraindications_by_drug_name, FDA_get_warnings_by_drug_name*

Phase 4: Mechanism-Based Adverse Event Context

4.1 Target Safety Profile

# Get target safety data from OpenTargets
# First get target ensembl ID from MOA result
target_id = "ENSG00000113161"  # HMGCR from Phase 0

safety = tu.tools.OpenTargets_get_target_safety_profile_by_ensemblID(ensemblId=target_id)
# Response: {data: {target: {id: "...", approvedSymbol: "HMGCR",
#   safetyLiabilities: [{event: "Decrease, Fertility", eventId: "...",
#     effects: [{direction: "Inhibition/Decrease/Downregulation"}],
#     studies: [{type: "cell-based"}], datasource: "AOP-Wiki"}]}}}

# Get OpenTargets adverse events (uses FAERS data)
ot_aes = tu.tools.OpenTargets_get_drug_adverse_events_by_chemblId(chemblId="CHEMBL1487")
# Response: {data: {drug: {adverseEvents: {count: 13, criticalValue: 513.67,
#   rows: [{name: "myalgia", meddraCode: "10028411", count: 4126, logLR: 6067.33}, ...]}}}}

4.2 ADMET Predictions (if SMILES available)

# Get SMILES from DrugBank/PharmGKB
smiles = "CC(C)C1=C(C(=C(N1CC[C@H](C[C@H](CC(=O)O)O)O)C2=CC=C(C=C2)F)C3=CC=CC=C3)C(=O)NC4=CC=CC=C4"

# Toxicity predictions
toxicity = tu.tools.ADMETAI_predict_toxicity(smiles=[smiles])
# Response: predictions for hepatotoxicity, cardiotoxicity, etc.

# CYP interaction predictions
cyp = tu.tools.ADMETAI_predict_CYP_interactions(smiles=[smiles])
# Response: CYP inhibition/substrate predictions

4.3 Drug Warnings from OpenTargets

# Drug warnings (withdrawals, safety warnings)
warnings = tu.tools.OpenTargets_get_drug_warnings_by_chemblId(chemblId="CHEMBL1487")
# Response: {data: {drug: {id: "CHEMBL1487", name: "ATORVASTATIN"}}}
# Note: Empty if no warnings exist

4.4 Output for Report

## 5. Mechanism-Based Adverse Event Context

### 5.1 Target Safety Profile (HMGCR)
| Safety Liability | Direction | Evidence | Source |
|-----------------|-----------|----------|--------|
| Decreased fertility | Inhibition | Cell-based | AOP-Wiki |

### 5.2 OpenTargets Significant Adverse Events
| Adverse Event | FAERS Count | log(LR) | MedDRA Code |
|---------------|-------------|---------|-------------|
| Myalgia | 4,126 | 6,067 | 10028411 |
| Rhabdomyolysis | 2,546 | 4,788 | 10039020 |
| CPK increased | 1,709 | 2,909 | 10005470 |

### 5.3 ADMET Predictions
| Property | Prediction | Confidence |
|----------|-----------|------------|
| Hepatotoxicity | Moderate risk | 0.65 |
| Cardiotoxicity (hERG) | Low risk | 0.23 |
| CYP3A4 substrate | Yes | 0.92 |

*Source: OpenTargets, ADMETAI*

Phase 5: Comparative Safety Analysis

5.1 Compare to Drug Class

# Head-to-head comparison with class member
comparison = tu.tools.FAERS_compare_drugs(
    operation="compare_drugs",
    drug1="ATORVASTATIN",
    drug2="SIMVASTATIN",
    adverse_event="Rhabdomyolysis"
)
# Response: {status: "success", adverse_event: "Rhabdomyolysis",
#   drug1: {name: "ATORVASTATIN", metrics: {PRR: {value: 4.79, ...}, ROR: {...}, IC: {...}},
#           signal_detection: {signal_detected: true, signal_strength: "Strong signal"}},
#   drug2: {name: "SIMVASTATIN", metrics: {PRR: {value: 12.78, ...}, ...}},
#   comparison: "SIMVASTATIN shows stronger signal than ATORVASTATIN"}

# Compare multiple events across class members
class_drugs = ["ATORVASTATIN", "SIMVASTATIN", "ROSUVASTATIN", "PRAVASTATIN"]
key_events = ["Rhabdomyolysis", "Myalgia", "Hepatotoxicity", "Diabetes mellitus"]
# Run FAERS_compare_drugs for each pair and event combination

# Aggregate adverse events across drug class
class_aes = tu.tools.FAERS_count_additive_adverse_reactions(
    medicinalproducts=class_drugs
)
# Response: [{term: "FATIGUE", count: N}, ...]

# Aggregate seriousness across class
class_serious = tu.tools.FAERS_count_additive_seriousness_classification(
    medicinalproducts=class_drugs
)
# Response: [{term: "Serious", count: N}, {term: "Non-serious", count: N}]

5.2 Output for Report

## 6. Comparative Safety Analysis (Statin Class)

### 6.1 Head-to-Head: Rhabdomyolysis

| Drug | PRR | PRR 95% CI | ROR | Cases | Signal |
|------|-----|------------|-----|-------|--------|
| Simvastatin | 12.78 | 12.43-13.14 | 13.05 | 5,234 | **STRONG** |
| Atorvastatin | 4.79 | 4.59-5.00 | 4.83 | 2,226 | **STRONG** |
| Rosuvastatin | 3.45 | 3.21-3.71 | 3.47 | 1,102 | Moderate |
| Pravastatin | 5.67 | 5.28-6.09 | 5.72 | 1,876 | **STRONG** |

### 6.2 Class-Wide vs Drug-Specific Signals

| Signal Type | Events |
|-------------|--------|
| **Class-wide** (all statins) | Myalgia, Rhabdomyolysis, CPK elevation, Hepatotoxicity |
| **Drug-specific** (atorvastatin) | [None identified - all signals are class-wide] |

*Source: FAERS via FAERS_compare_drugs*

Phase 6: Drug-Drug Interactions & Risk Factors

6.1 Drug-Drug Interactions

# FDA label DDIs
ddi_label = tu.tools.FDA_get_drug_interactions_by_drug_name(drug_name="atorvastatin")
# Response: {results: [{drug_interactions: ["...text..."]}]}

# DrugBank interactions
ddi_db = tu.tools.drugbank_get_drug_interactions_by_drug_name_or_id(
    query="atorvastatin", case_sensitive=False, exact_match=False, limit=3
)

# DailyMed DDIs
ddi_dailymed = tu.tools.DailyMed_parse_drug_interactions(drug_name="atorvastatin")

6.2 Pharmacogenomic Risk Factors

# PharmGKB drug search
pgx_search = tu.tools.PharmGKB_search_drugs(query="atorvastatin")
# Response: {status: "success", data: [{id: "PA448500", name: "atorvastatin", smiles: "..."}]}

# Get detailed PGx info
pgx_details = tu.tools.PharmGKB_get_drug_details(drug_id="PA448500")

# PharmGKB dosing guidelines
dosing = tu.tools.PharmGKB_get_dosing_guidelines(gene="SLCO1B1")
# SLCO1B1 is key pharmacogene for statins

# FDA PGx biomarkers
fda_pgx = tu.tools.fda_pharmacogenomic_biomarkers(drug_name="atorvastatin", limit=10)
# Response: {count: N, results: [{drug_name: "...", biomarker: "...", ...}]}
# Note: May return empty results for some drugs

6.3 Output for Report

## 7. Drug-Drug Interactions & Pharmacogenomic Risk

### 7.1 Key Drug-Drug Interactions
| Interacting Drug | Mechanism | Risk | Management |
|-----------------|-----------|------|------------|
| Cyclosporine | CYP3A4 inhibition | Rhabdomyolysis | Avoid combination |
| Clarithromycin | CYP3A4 inhibition | Myopathy | Limit to 20mg/day |
| Gemfibrozil | Glucuronidation inhibition | Myopathy | Avoid combination |
| Niacin (>1g/day) | Additive myotoxicity | Myopathy | Monitor closely |

### 7.2 Pharmacogenomic Risk Factors
| Gene | Variant | Phenotype | Recommendation | Evidence |
|------|---------|-----------|----------------|----------|
| SLCO1B1 | rs4149056 (*5) | Reduced transport | Consider lower dose | Level 1A |
| CYP3A4 | *22 (rs35599367) | Poor metabolizer | Increased exposure | Level 3 |

*Source: FDA label, PharmGKB, fda_pharmacogenomic_biomarkers*

Phase 7: Literature Evidence

7.1 Search Published Literature

# PubMed safety studies
pubmed = tu.tools.PubMed_search_articles(
    query='atorvastatin adverse events safety rhabdomyolysis',
    limit=20
)
# Response: [{pmid: "...", title: "...", authors: [...], journal: "...",
#   pub_date: "...", pub_year: "...", doi: "..."}]

# Citation analysis via OpenAlex
openalex = tu.tools.openalex_search_works(
    query="atorvastatin safety adverse events",
    limit=15
)

# Preprints via EuropePMC
preprints = tu.tools.EuropePMC_search_articles(
    query="atorvastatin safety signal",
    source="PPR",
    pageSize=10
)

7.2 Output for Report

## 8. Literature Evidence

### 8.1 Key Safety Publications
| PMID | Title | Year | Journal |
|------|-------|------|---------|
| 41657777 | Differential musculoskeletal outcome reporting... | 2026 | Front Pharmacol |
| ... | ... | ... | ... |

### 8.2 Evidence Summary
| Evidence Type | Count | Key Findings |
|---------------|-------|--------------|
| Meta-analyses | 5 | Statin myopathy 5-10%, rhabdomyolysis rare |
| RCTs | 12 | CV benefit outweighs muscle risk |
| Case reports | 23 | Severe rhabdomyolysis with CYP3A4 inhibitors |

*Source: PubMed, OpenAlex*

Phase 8: Risk Assessment & Safety Signal Score

8.1 Safety Signal Score Calculation (0-100)

The Safety Signal Score quantifies overall drug safety concern on a 0-100 scale (higher = more concern).

Component 1: FAERS Signal Strength (0-35 points)

If any signal has PRR >= 5 AND ROR lower CI >= 3: 35 points
If any signal has PRR 3-5 AND ROR lower CI 2-3: 20 points
If any signal has PRR 2-3 AND ROR lower CI 1-2: 10 points
If no signals detected: 0 points

Component 2: Serious Adverse Events (0-30 points)

Deaths reported with high count (>100): 30 points
Deaths reported with low count (1-100): 25 points
Life-threatening events: 20 points
Hospitalizations only: 15 points
Non-serious only: 0 points

Component 3: FDA Label Warnings (0-25 points)

Boxed warning present: 25 points
Drug withdrawn or restricted: 25 points
Contraindications present: 15 points
Warnings and precautions: 10 points
Adverse reactions only: 5 points
No label warnings: 0 points

Component 4: Literature Evidence (0-10 points)

Meta-analyses confirming safety signals: 10 points
Multiple RCTs with safety concerns: 7 points
Case reports/case series: 4 points
No published safety concerns: 0 points

Total Score Interpretation:

Score Range Interpretation Action
75-100 High concern Serious safety signals; requires immediate regulatory attention
50-74 Moderate concern Significant monitoring needed; consider risk mitigation
25-49 Low-moderate concern Routine enhanced monitoring; standard risk management
0-24 Low concern Standard safety profile; routine pharmacovigilance

8.2 Evidence Grading

Tier Criteria Example
T1 Boxed warning, confirmed by RCTs, PRR > 10 Metformin: Lactic acidosis
T2 Label warning + FAERS signal (PRR 3-10) + published studies Atorvastatin: Rhabdomyolysis
T3 FAERS signal (PRR 2-3) + case reports Atorvastatin: Pancreatitis
T4 Computational prediction only (ADMET) or weak signal ADMETAI hepatotoxicity prediction

8.3 Output for Report

## 9. Risk Assessment

### 9.1 Safety Signal Score: 62/100 (MODERATE CONCERN)

| Component | Score | Max | Rationale |
|-----------|-------|-----|-----------|
| FAERS Signal Strength | 35 | 35 | Strong signals (PRR >= 5 for rhabdomyolysis) |
| Serious Adverse Events | 15 | 30 | Hospitalizations; deaths uncommon for drug itself |
| FDA Label Warnings | 10 | 25 | Warnings/precautions but no boxed warning |
| Literature Evidence | 7 | 10 | Multiple RCTs confirm muscle-related risks |
| **TOTAL** | **62** | **100** | **MODERATE CONCERN** |

### 9.2 Evidence-Graded Signals

| Signal | Grade | PRR | Serious | Label | Literature | Overall |
|--------|-------|-----|---------|-------|------------|---------|
| Rhabdomyolysis | T2 | 4.79 | Yes | Warning | Confirmed | Moderate |
| Myopathy | T2 | 6.12 | Yes | Warning | Confirmed | Moderate |
| Hepatotoxicity | T3 | 3.45 | Rare | Warning | Case reports | Low-Moderate |
| Diabetes risk | T3 | 1.89 | No | Warning | RCT data | Low |

Phase 9: Report Synthesis & Recommendations

9.1 Report Template

File: [DRUG]_adverse_event_report.md

# Adverse Drug Event Signal Detection Report: [DRUG]

**Generated**: [Date] | **Drug**: [Generic Name] | **ChEMBL ID**: [ID]
**Safety Signal Score**: [XX/100] ([INTERPRETATION])

---

## Executive Summary

[2-3 paragraph summary of key findings]

**Key Safety Signals**:
1. [Strongest signal with PRR/ROR]
2. [Second signal]
3. [Third signal]

**Regulatory Status**: [Boxed warning Y/N] | [Withdrawn Y/N] | [Restrictions]

---

## 1. Drug Identification
[Phase 0 output]

## 2. FAERS Adverse Event Profile
[Phase 1 output]

## 3. Disproportionality Analysis
[Phase 2 output]

## 4. FDA Label Safety Information
[Phase 3 output]

## 5. Mechanism-Based Context
[Phase 4 output]

## 6. Comparative Safety Analysis
[Phase 5 output]

## 7. Drug-Drug Interactions & PGx Risk
[Phase 6 output]

## 8. Literature Evidence
[Phase 7 output]

## 9. Risk Assessment
[Phase 8 output]

## 10. Clinical Recommendations

### 10.1 Monitoring Recommendations
| Parameter | Frequency | Rationale |
|-----------|-----------|-----------|
| [Lab test] | [Frequency] | [Why] |

### 10.2 Risk Mitigation Strategies
| Risk | Mitigation | Evidence |
|------|-----------|----------|
| [Risk] | [Strategy] | [Source] |

### 10.3 Patient Counseling Points
- [Point 1]
- [Point 2]

### 10.4 Populations at Higher Risk
| Population | Risk Factor | Recommendation |
|-----------|-------------|----------------|
| [Group] | [Factor] | [Action] |

---

## 11. Completeness Checklist
[See below]

## 12. Data Sources
[All tools and databases used with timestamps]

Completeness Checklist

Phase 0: Drug Disambiguation

  • Generic name resolved
  • ChEMBL ID obtained
  • DrugBank ID obtained
  • Drug class identified
  • Mechanism of action stated
  • Primary target identified
  • Blackbox/withdrawal status checked

Phase 1: FAERS Profiling

  • Top adverse events queried (>=15 events)
  • Seriousness distribution obtained
  • Outcome distribution obtained
  • Age distribution obtained
  • Death-related events counted
  • Reporter country distribution obtained

Phase 2: Disproportionality Analysis

  • PRR calculated for >= 10 adverse events
  • ROR with 95% CI for each event
  • IC with 95% CI for each event
  • Signal strength classified for each
  • Demographics stratified for strong signals

Phase 3: FDA Label

  • Boxed warnings checked (or confirmed none)
  • Contraindications extracted
  • Warnings and precautions extracted
  • Adverse reactions from label
  • Drug interactions from label
  • Special populations (pregnancy, geriatric, pediatric)

Phase 4: Mechanism Context

  • Target safety profile (OpenTargets)
  • OpenTargets adverse events queried
  • ADMET predictions (if SMILES available)

Phase 5: Comparative Analysis

  • At least 1 class comparison performed
  • Class-wide vs drug-specific signals identified
  • Aggregate class AEs computed (if applicable)

Phase 6: DDIs & PGx

  • DDIs from FDA label extracted
  • PharmGKB queried
  • Dosing guidelines checked
  • FDA PGx biomarkers checked

Phase 7: Literature

  • PubMed searched (>=10 articles)
  • OpenAlex citation analysis (if time permits)
  • Key safety publications cited

Phase 8: Risk Assessment

  • Safety Signal Score calculated (0-100)
  • Each signal evidence-graded (T1-T4)
  • Score interpretation provided

Phase 9: Report

  • Report file created and saved
  • Executive summary written
  • Monitoring recommendations provided
  • Risk mitigation strategies listed
  • Patient counseling points included
  • All sources cited

Tool Parameter Reference (Verified)

FAERS Tools (OpenFDA-based)

Tool Key Parameters Notes
FAERS_count_reactions_by_drug_event medicinalproduct (REQUIRED), patientsex, patientagegroup, occurcountry Returns [{term, count}]
FAERS_count_seriousness_by_drug_event medicinalproduct (REQUIRED), patientsex, patientagegroup, occurcountry Returns [{term: "Serious"/"Non-serious", count}]
FAERS_count_outcomes_by_drug_event medicinalproduct (REQUIRED), patientsex, patientagegroup, occurcountry Returns [{term: "Fatal"/"Recovered"/..., count}]
FAERS_count_patient_age_distribution medicinalproduct (REQUIRED) Returns [{term: "Elderly"/"Adult"/..., count}]
FAERS_count_death_related_by_drug medicinalproduct (REQUIRED) Returns [{term: "alive"/"death", count}]
FAERS_count_reportercountry_by_drug_event medicinalproduct (REQUIRED), patientsex, patientagegroup, serious Returns [{term: "US"/"GB"/..., count}]
FAERS_search_adverse_event_reports medicinalproduct, limit (max 100), skip Returns individual case reports with patient/drug/reaction data
FAERS_search_reports_by_drug_and_reaction medicinalproduct (REQUIRED), reactionmeddrapt (REQUIRED), limit, skip, patientsex, serious Returns individual reports filtered by specific reaction
FAERS_search_serious_reports_by_drug medicinalproduct (REQUIRED), seriousnessdeath, seriousnesshospitalization, seriousnesslifethreatening, seriousnessdisabling, limit Returns serious event reports

FAERS Analytics Tools (operation-based)

Tool Key Parameters Notes
FAERS_calculate_disproportionality operation="calculate_disproportionality", drug_name (REQUIRED), adverse_event (REQUIRED) Returns PRR, ROR, IC with 95% CI and signal detection
FAERS_analyze_temporal_trends operation="analyze_temporal_trends", drug_name (REQUIRED), adverse_event (optional) Returns yearly counts and trend direction
FAERS_compare_drugs operation="compare_drugs", drug1 (REQUIRED), drug2 (REQUIRED), adverse_event (REQUIRED) Returns PRR/ROR/IC for both drugs side-by-side
FAERS_filter_serious_events operation="filter_serious_events", drug_name (REQUIRED), seriousness_type (death/hospitalization/disability/life_threatening/all) Returns top serious reactions with counts
FAERS_stratify_by_demographics operation="stratify_by_demographics", drug_name (REQUIRED), adverse_event (REQUIRED), stratify_by (sex/age/country) Returns stratified counts and percentages. Sex codes: 0=Unknown, 1=Male, 2=Female
FAERS_rollup_meddra_hierarchy operation="rollup_meddra_hierarchy", drug_name (REQUIRED) Returns top 50 preferred terms with counts

FAERS Aggregate Tools (multi-drug)

Tool Key Parameters Notes
FAERS_count_additive_adverse_reactions medicinalproducts (REQUIRED, array), patientsex, patientagegroup, occurcountry, serious, seriousnessdeath Aggregates AE counts across multiple drugs
FAERS_count_additive_seriousness_classification medicinalproducts (REQUIRED, array), patientsex, patientagegroup, occurcountry Aggregates seriousness across multiple drugs
FAERS_count_additive_reaction_outcomes medicinalproducts (REQUIRED, array) Aggregates outcomes across multiple drugs

FDA Label Tools

Tool Key Parameters Notes
FDA_get_boxed_warning_info_by_drug_name drug_name Returns {error: {code: "NOT_FOUND"}} if no boxed warning
FDA_get_contraindications_by_drug_name drug_name Returns {meta: {total: N}, results: [{contraindications: [...]}]}
FDA_get_adverse_reactions_by_drug_name drug_name Returns {meta: {total: N}, results: [{adverse_reactions: [...]}]}
FDA_get_warnings_by_drug_name drug_name Returns {meta: {total: N}, results: [{warnings: [...]}]}
FDA_get_drug_interactions_by_drug_name drug_name Returns {meta: {total: N}, results: [{drug_interactions: [...]}]}
FDA_get_pharmacogenomics_info_by_drug_name drug_name Returns PGx info from label
FDA_get_pregnancy_or_breastfeeding_info_by_drug_name drug_name Returns pregnancy info
FDA_get_geriatric_use_info_by_drug_name drug_name Returns geriatric use info
FDA_get_pediatric_use_info_by_drug_name drug_name Returns pediatric info

OpenTargets Tools

Tool Key Parameters Notes
OpenTargets_get_drug_chembId_by_generic_name drugName Returns {data: {search: {hits: [{id, name, description}]}}}
OpenTargets_get_drug_adverse_events_by_chemblId chemblId Returns {data: {drug: {adverseEvents: {count, rows: [{name, meddraCode, count, logLR}]}}}}
OpenTargets_get_drug_blackbox_status_by_chembl_ID chemblId Returns {data: {drug: {hasBeenWithdrawn, blackBoxWarning}}}
OpenTargets_get_drug_warnings_by_chemblId chemblId Returns drug warnings (may be empty)
OpenTargets_get_drug_mechanisms_of_action_by_chemblId chemblId Returns {data: {drug: {mechanismsOfAction: {rows: [{mechanismOfAction, actionType, targetName, targets}]}}}}
OpenTargets_get_drug_indications_by_chemblId chemblId Returns approved and investigational indications
OpenTargets_get_target_safety_profile_by_ensemblID ensemblId Returns {data: {target: {safetyLiabilities: [{event, effects, studies, datasource}]}}}

DrugBank Tools

Tool Key Parameters Notes
drugbank_get_safety_by_drug_name_or_drugbank_id query, case_sensitive (bool), exact_match (bool), limit Returns toxicity, food interactions
drugbank_get_targets_by_drug_name_or_drugbank_id query, case_sensitive, exact_match, limit Returns drug targets
drugbank_get_drug_interactions_by_drug_name_or_id query, case_sensitive, exact_match, limit Returns DDIs
drugbank_get_pharmacology_by_drug_name_or_drugbank_id query, case_sensitive, exact_match, limit Returns pharmacology

PharmGKB Tools

Tool Key Parameters Notes
PharmGKB_search_drugs query Returns {data: [{id, name, smiles}]}
PharmGKB_get_drug_details drug_id (e.g., "PA448500") Returns detailed drug info
PharmGKB_get_dosing_guidelines guideline_id, gene (both optional) Returns dosing guidelines
PharmGKB_get_clinical_annotations annotation_id, gene_id (both optional) Returns clinical annotations
fda_pharmacogenomic_biomarkers drug_name, biomarker, limit Returns {count, results: [...]}

ADMETAI Tools

Tool Key Parameters Notes
ADMETAI_predict_toxicity smiles (REQUIRED, array of strings) Predicts hepatotoxicity, cardiotoxicity, etc.
ADMETAI_predict_CYP_interactions smiles (REQUIRED, array) Predicts CYP inhibition/substrate

Literature Tools

Tool Key Parameters Notes
PubMed_search_articles query, limit Returns list of article dicts
openalex_search_works query, limit Returns works with citation counts
EuropePMC_search_articles query, source ("PPR" for preprints), pageSize Returns articles including preprints
search_clinical_trials query_term (REQUIRED), condition, intervention, pageSize Returns clinical trials

Fallback Chains

Primary Tool Fallback 1 Fallback 2
FAERS_calculate_disproportionality Manual calculation from FAERS_count_* data Literature PRR values
FAERS_count_reactions_by_drug_event FAERS_rollup_meddra_hierarchy OpenTargets adverse events
FDA_get_boxed_warning_info_by_drug_name OpenTargets_get_drug_blackbox_status_by_chembl_ID DrugBank safety
FDA_get_contraindications_by_drug_name FDA_get_warnings_by_drug_name DrugBank safety
OpenTargets_get_drug_chembId_by_generic_name ChEMBL_search_drugs Manual search
PharmGKB_search_drugs fda_pharmacogenomic_biomarkers FDA label PGx section
PubMed_search_articles openalex_search_works EuropePMC_search_articles

Common Patterns

Pattern 1: Full Safety Signal Profile for a Single Drug

Use all phases (0-9) for comprehensive report. Best for regulatory submissions, safety reviews.

Pattern 2: Specific Adverse Event Investigation

Focus on Phases 0, 2, 3, 7. User asks "Does [drug] cause [event]?" - calculate disproportionality for that specific event, check label, search literature.

Pattern 3: Drug Class Comparison

Focus on Phases 0, 2, 5. Compare 3-5 drugs in same class for a specific adverse event using FAERS_compare_drugs.

Pattern 4: Emerging Signal Detection

Focus on Phases 1, 2, 7. Screen top 20+ FAERS events for signals, identify any not in FDA label (Phase 3), search recent literature for confirmation.

Pattern 5: Pharmacogenomic Risk Assessment

Focus on Phases 0, 6. Identify genetic risk factors for adverse events using PharmGKB and FDA PGx biomarkers.

Pattern 6: Pre-Approval Safety Assessment

Focus on Phases 4, 7. Use ADMET predictions and target safety profiles when FAERS data is limited (new drugs).

Edge Cases

Drug with No FAERS Reports

  • Skip Phases 1-2
  • Rely on FDA label (Phase 3), mechanism predictions (Phase 4), and literature (Phase 7)
  • Safety Signal Score will be lower due to lack of signal detection data

Generic vs Brand Name

  • Always try both names in FAERS queries (FAERS uses brand names sometimes)
  • Use OpenTargets_get_drug_chembId_by_generic_name to resolve to standard identifier
  • Use FDA_get_brand_name_generic_name for name cross-reference

Drug Combinations

  • Use FAERS_search_reports_by_drug_combination for polypharmacy analysis
  • Distinguish combination AEs from individual drug AEs
  • Use FAERS_count_additive_adverse_reactions for aggregate class analysis

Confounding by Indication

  • Compare AE profile to the disease being treated
  • Example: "Death" reports for chemotherapy drugs may reflect disease progression
  • Always note this limitation in the report

Drugs with Boxed Warnings

  • Score component automatically 25/25 for label warnings
  • Prioritize boxed warning events in disproportionality analysis
  • Cross-reference boxed warning with FAERS signal strength
Related skills

More from wu-yc/labclaw

Installs
14
Repository
wu-yc/labclaw
GitHub Stars
981
First Seen
Mar 13, 2026
Security Audits
Gen Agent Trust HubPass
SocketPass
SnykWarn