Back to Details

tooluniverse-chemical-safety

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Chemical Safety & Toxicology Assessment

化学安全与毒理学评估

Comprehensive chemical safety and toxicology analysis integrating predictive AI models, curated toxicogenomics databases, regulatory safety data, and chemical-biological interaction networks. Generates structured risk assessment reports with evidence grading.
整合预测AI模型、经过整理的毒理基因组学数据库、监管安全数据以及化学生物相互作用网络的综合性化学安全与毒理学分析。可生成带有证据分级的结构化风险评估报告。

When to Use This Skill

何时使用该技能

Triggers:
  • "Is this chemical toxic?" / "What are the toxicity endpoints for [compound]?"
  • "Assess the safety profile of [drug/chemical]"
  • "What are the ADMET properties of [SMILES]?"
  • "What genes does [chemical] interact with?"
  • "What diseases are linked to [chemical] exposure?"
  • "Predict toxicity for these molecules"
  • "Drug safety assessment for [drug name]"
  • "Environmental health risk of [chemical]"
  • "Chemical hazard profiling"
  • "Toxicogenomic analysis of [compound]"
Use Cases:
  1. Predictive Toxicology: AI-predicted toxicity endpoints (AMES mutagenicity, DILI, LD50, carcinogenicity, skin reactions) for novel compounds via SMILES
  2. ADMET Profiling: Full absorption, distribution, metabolism, excretion, toxicity characterization
  3. Toxicogenomics: Chemical-gene interaction mapping, gene-disease associations from CTD
  4. Regulatory Safety: FDA label warnings, boxed warnings, contraindications, adverse reactions
  5. Drug Safety Assessment: Combined DrugBank safety + FDA labels + adverse event data
  6. Chemical-Protein Interactions: STITCH-based chemical-protein binding and interaction networks
  7. Environmental Toxicology: Chemical-disease associations for environmental contaminants
触发场景:
  • "该化学物质有毒吗？" / "[化合物]的毒性终点是什么？"
  • "评估[药物/化学物质]的安全档案"
  • "[SMILES]的ADMET特性是什么？"
  • "[化学物质]与哪些基因相互作用？"
  • "接触[化学物质]会关联哪些疾病？"
  • "预测这些分子的毒性"
  • "[药物名称]的药物安全评估"
  • "[化学物质]的环境健康风险"
  • "化学危害分析"
  • "[化合物]的毒理基因组学分析"
适用场景:
  1. 预测毒理学: 通过SMILES对新型化合物进行AI预测的毒性终点分析（AMES致突变性、DILI肝毒性、LD50半数致死量、致癌性、皮肤反应）
  2. ADMET分析: 完整的吸收、分布、代谢、排泄、毒性特征分析
  3. 毒理基因组学: 化学-基因相互作用映射、CTD数据库中的基因-疾病关联分析
  4. 监管安全: FDA标签警告、黑框警告、禁忌症、不良反应
  5. 药物安全评估: 结合DrugBank安全数据、FDA标签数据及不良事件数据
  6. 化学-蛋白质相互作用: 基于STITCH的化学-蛋白质结合及相互作用网络分析
  7. 环境毒理学: 环境污染物的化学-疾病关联分析

KEY PRINCIPLES

核心原则

  1. Report-first approach - Create report file FIRST, then populate progressively
  2. Tool parameter verification - Verify params via 
    get_tool_info
    before calling unfamiliar tools
  3. Evidence grading - Grade all safety claims by evidence strength (T1-T4)
  4. Citation requirements - Every toxicity finding must have inline source attribution
  5. Mandatory completeness - All sections must exist with data minimums or explicit "No data" notes
  6. Disambiguation first - Resolve compound identity (name -> SMILES, CID, ChEMBL ID) before analysis
  7. Negative results documented - "No toxicity signals found" is data; empty sections are failures
  8. Conservative risk assessment - When evidence is ambiguous, flag as "requires further investigation"
  9. English-first queries - Always use English chemical/drug names in tool calls
  1. 报告优先原则 - 先创建报告文件，再逐步填充内容
  2. 工具参数验证 - 在调用不熟悉的工具前，通过
    get_tool_info
    验证参数
  3. 证据分级 - 所有安全结论需按证据强度分级（T1-T4）
  4. 引用要求 - 每一项毒性发现必须有内联来源标注
  5. 完整性要求 - 所有章节必须存在，至少包含基础数据或明确标注“无数据”
  6. 先消歧义 - 在分析前先确认化合物身份（名称→SMILES、CID、ChEMBL ID）
  7. 记录阴性结果 - “未发现毒性信号”属于有效数据；空章节视为失败
  8. 保守风险评估 - 当证据不明确时，标记为“需进一步研究”
  9. 英文优先查询 - 在工具调用中始终使用英文化学/药物名称

Evidence Grading System (MANDATORY)

证据分级系统（强制要求）

Grade every toxicity claim by evidence strength:
TierSymbolCriteriaExamples
T1[T1]Direct human evidence, regulatory findingFDA boxed warning, clinical trial toxicity, human case reports
T2[T2]Animal studies, validated in vitroNonclinical toxicology, AMES positive, animal LD50
T3[T3]Computational prediction, association dataADMET-AI prediction, CTD association, QSAR model
T4[T4]Database annotation, text-minedLiterature mention, database entry without validation
所有毒性结论需按证据强度分级：
等级标识标准示例
T1[T1]直接人体证据、监管机构结论FDA黑框警告、临床试验毒性、人体病例报告
T2[T2]动物研究、已验证的体外实验非临床毒理学、AMES阳性、动物LD50数据
T3[T3]计算预测、关联数据ADMET-AI预测、CTD关联、QSAR模型
T4[T4]数据库注释、文本挖掘结果文献提及、未经验证的数据库条目

Required Evidence Grading Locations

证据分级必填位置

Evidence grades MUST appear in:
  1. Executive Summary - Key toxicity findings graded
  2. Toxicity Predictions - Every ADMET-AI endpoint with confidence note
  3. Regulatory Safety - FDA findings marked [T1]
  4. Chemical-Gene Interactions - CTD data marked by curation status
  5. Risk Assessment - Final risk classification with supporting evidence tiers
证据分级必须出现在：
  1. 执行摘要 - 关键毒性发现需分级
  2. 毒性预测 - 每个ADMET-AI终点需标注置信度
  3. 监管安全 - FDA发现标记为[T1]
  4. 化学-基因相互作用 - CTD数据需标注整理状态
  5. 风险评估 - 最终风险分类需附带支持证据等级

Core Strategy: 8 Research Dimensions

核心策略：8个研究维度

Chemical/Drug Query
|
+-- PHASE 0: Compound Disambiguation (ALWAYS FIRST)
|   +-- Resolve name -> SMILES, PubChem CID, ChEMBL ID
|   +-- Get molecular formula, weight, canonical structure
|
+-- PHASE 1: Predictive Toxicology (ADMET-AI)
|   +-- Mutagenicity (AMES)
|   +-- Hepatotoxicity (DILI, ClinTox)
|   +-- Carcinogenicity
|   +-- Acute toxicity (LD50)
|   +-- Skin reactions
|   +-- Stress response pathways
|   +-- Nuclear receptor activity
|
+-- PHASE 2: ADMET Properties
|   +-- Absorption: BBB penetrance, bioavailability
|   +-- Distribution: clearance, volume of distribution
|   +-- Metabolism: CYP interactions (1A2, 2C9, 2C19, 2D6, 3A4)
|   +-- Physicochemical: solubility, lipophilicity, pKa
|
+-- PHASE 3: Toxicogenomics (CTD)
|   +-- Chemical-gene interactions
|   +-- Chemical-disease associations
|   +-- Affected biological pathways
|
+-- PHASE 4: Regulatory Safety (FDA Labels)
|   +-- Boxed warnings (Black Box)
|   +-- Contraindications
|   +-- Adverse reactions
|   +-- Warnings and precautions
|   +-- Nonclinical toxicology
|
+-- PHASE 5: Drug Safety Profile (DrugBank)
|   +-- Toxicity data
|   +-- Contraindications
|   +-- Drug interactions affecting safety
|
+-- PHASE 6: Chemical-Protein Interactions (STITCH)
|   +-- Direct chemical-protein binding
|   +-- Interaction confidence scores
|   +-- Off-target effects
|
+-- PHASE 7: Structural Alerts (ChEMBL)
|   +-- Known toxic substructures (PAINS, Brenk)
|   +-- Structural alert flags
|
+-- SYNTHESIS: Integrated Risk Assessment
    +-- Aggregate all evidence tiers
    +-- Risk classification (Low/Medium/High/Critical)
    +-- Data gaps and recommendations
Chemical/Drug Query
|
+-- PHASE 0: Compound Disambiguation (ALWAYS FIRST)
|   +-- Resolve name -> SMILES, PubChem CID, ChEMBL ID
|   +-- Get molecular formula, weight, canonical structure
|
+-- PHASE 1: Predictive Toxicology (ADMET-AI)
|   +-- Mutagenicity (AMES)
|   +-- Hepatotoxicity (DILI, ClinTox)
|   +-- Carcinogenicity
|   +-- Acute toxicity (LD50)
|   +-- Skin reactions
|   +-- Stress response pathways
|   +-- Nuclear receptor activity
|
+-- PHASE 2: ADMET Properties
|   +-- Absorption: BBB penetrance, bioavailability
|   +-- Distribution: clearance, volume of distribution
|   +-- Metabolism: CYP interactions (1A2, 2C9, 2C19, 2D6, 3A4)
|   +-- Physicochemical: solubility, lipophilicity, pKa
|
+-- PHASE 3: Toxicogenomics (CTD)
|   +-- Chemical-gene interactions
|   +-- Chemical-disease associations
|   +-- Affected biological pathways
|
+-- PHASE 4: Regulatory Safety (FDA Labels)
|   +-- Boxed warnings (Black Box)
|   +-- Contraindications
|   +-- Adverse reactions
|   +-- Warnings and precautions
|   +-- Nonclinical toxicology
|
+-- PHASE 5: Drug Safety Profile (DrugBank)
|   +-- Toxicity data
|   +-- Contraindications
|   +-- Drug interactions affecting safety
|
+-- PHASE 6: Chemical-Protein Interactions (STITCH)
|   +-- Direct chemical-protein binding
|   +-- Interaction confidence scores
|   +-- Off-target effects
|
+-- PHASE 7: Structural Alerts (ChEMBL)
|   +-- Known toxic substructures (PAINS, Brenk)
|   +-- Structural alert flags
|
+-- SYNTHESIS: Integrated Risk Assessment
    +-- Aggregate all evidence tiers
    +-- Risk classification (Low/Medium/High/Critical)
    +-- Data gaps and recommendations

Phase 0: Compound Disambiguation (ALWAYS FIRST)

阶段0：化合物消歧义（必须优先执行）

CRITICAL: Resolve compound identity before any analysis.
关键：在任何分析前先确认化合物身份。

Input Types Handled

支持的输入类型

Input FormatResolution Strategy
Drug name (e.g., "Aspirin")PubChem_get_CID_by_compound_name -> get SMILES from properties
SMILES stringUse directly for ADMET-AI; resolve to CID for other tools
PubChem CIDPubChem_get_compound_properties_by_CID -> get SMILES + name
ChEMBL IDChEMBL_get_molecule -> get SMILES + properties
输入格式解析策略
药物名称（如“Aspirin”）PubChem_get_CID_by_compound_name → 从属性中获取SMILES
SMILES字符串直接用于ADMET-AI；解析为CID供其他工具使用
PubChem CIDPubChem_get_compound_properties_by_CID → 获取SMILES + 名称
ChEMBL IDChEMBL_get_molecule → 获取SMILES + 属性

Resolution Steps

解析步骤

  1. Input detection: Determine if input is name, SMILES, CID, or ChEMBL ID
    • SMILES: contains typical SMILES characters (=, #, [, ], (, ), c, n, o and no spaces in middle)
    • CID: numeric only
    • ChEMBL: starts with "CHEMBL"
    • Otherwise: treat as compound name
  2. Name to CID: 
    PubChem_get_CID_by_compound_name(name=<compound_name>)
  3. CID to properties: 
    PubChem_get_compound_properties_by_CID(cid=<cid>)
  4. Extract SMILES: Get SMILES from PubChem properties (field: 
    ConnectivitySMILES
    , 
    CanonicalSMILES
    , or 
    IsomericSMILES
    depending on response format)
  5. Store resolved IDs: Maintain dict with 
    name
    , 
    smiles
    , 
    cid
    , 
    formula
    , 
    weight
    , 
    inchi
  1. 输入检测: 判断输入是名称、SMILES、CID还是ChEMBL ID
    • SMILES: 包含典型SMILES字符（=, #, [, ], (, ), c, n, o，且中间无空格）
    • CID: 仅包含数字
    • ChEMBL: 以“CHEMBL”开头
    • 其他情况: 视为化合物名称
  2. 名称转CID: 
    PubChem_get_CID_by_compound_name(name=<compound_name>)
  3. CID转属性: 
    PubChem_get_compound_properties_by_CID(cid=<cid>)
  4. 提取SMILES: 从PubChem属性中获取SMILES（字段: 
    ConnectivitySMILES
    、
    CanonicalSMILES
    或
    IsomericSMILES
    ，取决于返回格式）
  5. 存储解析后的ID: 维护包含
    name
    、
    smiles
    、
    cid
    、
    formula
    、
    weight
    、
    inchi
    的字典

Disambiguation Output

消歧输出

markdown
undefined
markdown
undefined

Compound Identity

化合物身份信息

PropertyValue
NameAcetaminophen
PubChem CID1983
SMILESCC(=O)Nc1ccc(O)cc1
FormulaC8H9NO2
Molecular Weight151.16
InChIInChI=1S/C8H9NO2/...

---
属性值
名称对乙酰氨基酚
PubChem CID1983
SMILESCC(=O)Nc1ccc(O)cc1
分子式C8H9NO2
分子量151.16
InChIInChI=1S/C8H9NO2/...

---

Phase 1: Predictive Toxicology (ADMET-AI)

阶段1：预测毒理学（ADMET-AI）

When: SMILES is available (from Phase 0 or provided directly)
Objective: Run comprehensive AI-predicted toxicity endpoints
适用场景: 已获取SMILES（来自阶段0或直接提供）
目标: 运行全面的AI预测毒性终点

Tools Used

使用的工具

All ADMET-AI tools take the same parameter format:
ToolPredicted EndpointsParameter
ADMETAI_predict_toxicity
AMES, Carcinogens_Lagunin, ClinTox, DILI, LD50_Zhu, Skin_Reaction, hERG
smiles
: list[str]
ADMETAI_predict_stress_response
Stress response pathway activation (ARE, ATAD5, HSE, MMP, p53)
smiles
: list[str]
ADMETAI_predict_nuclear_receptor_activity
AhR, AR, ER, PPARg, Aromatase nuclear receptor activity
smiles
: list[str]
所有ADMET-AI工具采用相同的参数格式:
工具预测终点参数
ADMETAI_predict_toxicity
AMES、Carcinogens_Lagunin、ClinTox、DILI、LD50_Zhu、Skin_Reaction、hERG
smiles
: list[str]
ADMETAI_predict_stress_response
应激反应通路激活（ARE、ATAD5、HSE、MMP、p53）
smiles
: list[str]
ADMETAI_predict_nuclear_receptor_activity
AhR、AR、ER、PPARg、芳香化酶核受体活性
smiles
: list[str]

Workflow

工作流程

  1. Call 
    ADMETAI_predict_toxicity(smiles=[resolved_smiles])
  2. Call 
    ADMETAI_predict_stress_response(smiles=[resolved_smiles])
  3. Call 
    ADMETAI_predict_nuclear_receptor_activity(smiles=[resolved_smiles])
  4. For each endpoint, interpret prediction:
    • Classification endpoints: Active (1) = toxic signal, Inactive (0) = no signal
    • Regression endpoints (LD50): Report numerical value with context
    • All predictions graded [T3] (computational prediction)
  1. 调用
    ADMETAI_predict_toxicity(smiles=[resolved_smiles])
  2. 调用
    ADMETAI_predict_stress_response(smiles=[resolved_smiles])
  3. 调用
    ADMETAI_predict_nuclear_receptor_activity(smiles=[resolved_smiles])
  4. 解读每个终点的预测结果:
    • 分类终点: 活性（1）= 有毒性信号，非活性（0）= 无毒性信号
    • 回归终点（LD50）: 报告数值及相关背景
    • 所有预测结果分级为[T3]（计算预测）

Decision Logic

决策逻辑

  • Multiple SMILES: Can batch up to ~10 SMILES in single call
  • Failed prediction: If ADMET-AI fails, note "prediction unavailable" (don't fail entire report)
  • Confidence: Note that AI predictions are [T3] evidence, not definitive
  • hERG flag: If hERG = Active, flag prominently (cardiac safety risk)
  • AMES flag: If AMES = Active, flag prominently (mutagenicity concern)
  • DILI flag: If DILI = Active, flag prominently (liver toxicity concern)
  • 多SMILES: 单次调用最多可批量处理约10个SMILES
  • 预测失败: 如果ADMET-AI调用失败，标注“预测不可用”（不终止整个报告）
  • 置信度: 需注明AI预测为[T3]证据，并非确定性结论
  • hERG标记: 如果hERG=活性，需突出标记（心脏安全风险）
  • AMES标记: 如果AMES=活性，需突出标记（致突变性风险）
  • DILI标记: 如果DILI=活性，需突出标记（肝毒性风险）

Output Table

输出表格

markdown
undefined
markdown
undefined

Toxicity Predictions [T3]

毒性预测结果 [T3]

EndpointPredictionInterpretationConcern Level
AMES MutagenicityInactiveNo mutagenic signalLow
CarcinogenicityInactiveNo carcinogenic signalLow
ClinToxActiveClinical toxicity signalHIGH
DILIActiveDrug-induced liver injury riskHIGH
LD50 (Zhu)2.45 log(mg/kg)~282 mg/kg (moderate)Medium
Skin ReactionInactiveNo skin sensitization signalLow
hERG InhibitionActiveCardiac arrhythmia riskHIGH
All predictions from ADMET-AI. Evidence tier: [T3] (computational prediction)

---
终点预测结果解读风险等级
AMES致突变性非活性无致突变信号低
致癌性非活性无致癌信号低
ClinTox临床毒性活性存在临床毒性信号高
DILI肝毒性活性存在药物诱导肝损伤风险高
LD50 (Zhu模型)2.45 log(mg/kg)~282 mg/kg（中等）中等
皮肤反应非活性无皮肤致敏信号低
hERG抑制活性存在心律失常风险高
所有预测结果来自ADMET-AI。证据等级: [T3]（计算预测）

---

Phase 2: ADMET Properties

阶段2：ADMET特性分析

When: SMILES is available
Objective: Full ADMET characterization beyond toxicity
适用场景: 已获取SMILES
目标: 完成毒性之外的完整ADMET特征分析

Tools Used

使用的工具

ToolProperties PredictedParameter
ADMETAI_predict_BBB_penetrance
Blood-brain barrier crossing probability
smiles
: list[str]
ADMETAI_predict_bioavailability
Oral bioavailability (F20%, F30%)
smiles
: list[str]
ADMETAI_predict_clearance_distribution
Clearance, VDss, half-life, PPB
smiles
: list[str]
ADMETAI_predict_CYP_interactions
CYP1A2, 2C9, 2C19, 2D6, 3A4 inhibition/substrate
smiles
: list[str]
ADMETAI_predict_physicochemical_properties
LogP, LogD, LogS, MW, pKa
smiles
: list[str]
ADMETAI_predict_solubility_lipophilicity_hydration
Aqueous solubility, lipophilicity, hydration free energy
smiles
: list[str]
工具预测特性参数
ADMETAI_predict_BBB_penetrance
血脑屏障穿透概率
smiles
: list[str]
ADMETAI_predict_bioavailability
口服生物利用度（F20%、F30%）
smiles
: list[str]
ADMETAI_predict_clearance_distribution
清除率、VDss稳态分布容积、半衰期、PPB血浆蛋白结合率
smiles
: list[str]
ADMETAI_predict_CYP_interactions
CYP1A2、2C9、2C19、2D6、3A4的抑制/底物特性
smiles
: list[str]
ADMETAI_predict_physicochemical_properties
LogP、LogD、LogS、分子量、pKa
smiles
: list[str]
ADMETAI_predict_solubility_lipophilicity_hydration
水溶性、亲脂性、水合自由能
smiles
: list[str]

Workflow

工作流程

  1. Call all 6 ADMET tools in parallel (independent calls)
  2. Compile results into Absorption / Distribution / Metabolism / Excretion sections
  3. Assess Lipinski Rule of 5 compliance from physicochemical properties
  4. Flag drug-drug interaction risks from CYP inhibition profiles
  1. 并行调用所有6个ADMET工具（独立调用）
  2. 将结果整理为吸收/分布/代谢/排泄章节
  3. 根据理化特性评估Lipinski五规则合规性
  4. 根据CYP抑制特征标记药物相互作用风险

Decision Logic

决策逻辑

  • BBB penetrant + toxicity: If BBB = Yes and any CNS toxicity endpoint active, flag as neurotoxicity risk
  • Low bioavailability: If F20% = Low, note absorption concerns
  • CYP inhibitor: If CYP3A4 inhibitor = Yes, flag high DDI risk
  • Lipinski violations: Count violations and report drug-likeness assessment
  • 血脑屏障穿透+毒性: 如果BBB=是且任何中枢神经系统毒性终点为活性，标记为神经毒性风险
  • 低生物利用度: 如果F20%=低，标注吸收问题
  • CYP抑制剂: 如果CYP3A4抑制剂=是，标记高药物相互作用风险
  • Lipinski规则违反: 统计违反次数并报告药物相似性评估结果

Output Format

输出格式

markdown
undefined
markdown
undefined

ADMET Profile [T3]

ADMET分析结果 [T3]

Absorption

吸收

PropertyValueInterpretation
BBB PenetranceYesCrosses blood-brain barrier
Bioavailability (F20%)85%Good oral absorption
属性值解读
血脑屏障穿透是可穿过血脑屏障
生物利用度（F20%）85%口服吸收良好

Distribution

分布

PropertyValueInterpretation
VDss1.2 L/kgModerate tissue distribution
PPB92%Highly protein bound
属性值解读
VDss稳态分布容积1.2 L/kg组织分布中等
PPB血浆蛋白结合率92%蛋白结合率高

Metabolism

代谢

CYP EnzymeSubstrateInhibitor
CYP1A2NoNo
CYP2C9YesNo
CYP2C19NoNo
CYP2D6NoNo
CYP3A4YesYes (DDI risk)
CYP酶底物抑制剂
CYP1A2否否
CYP2C9是否
CYP2C19否否
CYP2D6否否
CYP3A4是是（存在药物相互作用风险）

Excretion

排泄

PropertyValueInterpretation
Clearance8.5 mL/min/kgModerate clearance
Half-life6.2 hModerate half-life

---
属性值解读
清除率8.5 mL/min/kg清除率中等
半衰期6.2 h半衰期中等

---

Phase 3: Toxicogenomics (CTD)

阶段3：毒理基因组学（CTD）

When: Compound name is resolved
Objective: Map chemical-gene-disease relationships from curated CTD data
适用场景: 已解析化合物名称
目标: 从经过整理的CTD数据中映射化学-基因-疾病关系

Tools Used

使用的工具

ToolFunctionParameter
CTD_get_chemical_gene_interactions
Genes affected by chemical
input_terms
: str (chemical name)
CTD_get_chemical_diseases
Diseases linked to chemical exposure
input_terms
: str (chemical name)
工具功能参数
CTD_get_chemical_gene_interactions
受化学物质影响的基因
input_terms
: str（化学物质名称）
CTD_get_chemical_diseases
与化学物质暴露相关的疾病
input_terms
: str（化学物质名称）

Workflow

工作流程

  1. Call 
    CTD_get_chemical_gene_interactions(input_terms=compound_name)
  2. Call 
    CTD_get_chemical_diseases(input_terms=compound_name)
  3. Parse gene interactions: extract gene symbols, interaction types (increases/decreases expression, binding, etc.)
  4. Parse disease associations: extract disease names, evidence types (marker/mechanism/therapeutic)
  5. Identify most affected biological processes from gene list
  1. 调用
    CTD_get_chemical_gene_interactions(input_terms=compound_name)
  2. 调用
    CTD_get_chemical_diseases(input_terms=compound_name)
  3. 解析基因相互作用: 提取基因符号、相互作用类型（上调/下调表达、结合等）
  4. 解析疾病关联: 提取疾病名称、关联类型（标记/机制/治疗）
  5. 从基因列表中识别受影响最显著的生物通路

Decision Logic

决策逻辑

  • Direct evidence vs inferred: CTD separates curated direct evidence from inferred associations
  • Therapeutic vs toxic: Disease associations can be therapeutic (drug treats disease) or adverse (chemical causes disease)
  • Gene interaction types: Distinguish between expression changes, binding, and activity modulation
  • Prioritize marker/mechanism: These indicate stronger causal evidence than simple associations
  • Grade curated as [T2]: Direct curated CTD evidence from literature
  • Grade inferred as [T3]: Computationally inferred associations
  • 直接证据vs推断证据: CTD区分经过整理的直接证据与推断关联
  • 治疗vs毒性: 疾病关联可能是治疗性（药物治疗疾病）或不良性（化学物质导致疾病）
  • 基因相互作用类型: 区分表达变化、结合及活性调节
  • 优先标记/机制: 此类关联表示更强的因果证据，优于简单关联
  • 整理数据分级为[T2]: 来自文献的CTD直接整理证据
  • 推断数据分级为[T3]: 计算推断的关联

Output Format

输出格式

markdown
undefined
markdown
undefined

Toxicogenomics (CTD) [T2/T3]

毒理基因组学（CTD） [T2/T3]

Chemical-Gene Interactions (Top 20)

化学-基因相互作用（前20个）

GeneInteractionTypeEvidence
CYP1A2increases expressionmRNA[T2] curated
TP53affects activityprotein[T2] curated
............
Total interactions found: 156 Top affected pathways: Xenobiotic metabolism, Apoptosis, DNA damage response
基因相互作用类型证据
CYP1A2上调表达mRNA[T2] 经过整理
TP53影响活性蛋白质[T2] 经过整理
............
发现的相互作用总数: 156 受影响最显著的通路: 外源性物质代谢、细胞凋亡、DNA损伤应答

Chemical-Disease Associations (Top 10)

化学-疾病关联（前10个）

DiseaseAssociation TypeEvidence
Liver Neoplasmsmarker/mechanism[T2] curated
Contact Dermatitistherapeutic[T2] curated
.........

---
疾病关联类型证据
肝肿瘤标记/机制[T2] 经过整理
接触性皮炎治疗性[T2] 经过整理
.........

---

Phase 4: Regulatory Safety (FDA Labels)

阶段4：监管安全（FDA标签）

When: Compound has an approved drug name
Objective: Extract regulatory safety information from FDA drug labels
适用场景: 化合物为已获批药物名称
目标: 从FDA药物标签中提取监管安全信息

Tools Used

使用的工具

ToolInformation RetrievedParameter
FDA_get_boxed_warning_info_by_drug_name
Black box warnings (most serious)
drug_name
: str
FDA_get_contraindications_by_drug_name
Absolute contraindications
drug_name
: str
FDA_get_adverse_reactions_by_drug_name
Known adverse reactions
drug_name
: str
FDA_get_warnings_by_drug_name
Warnings and precautions
drug_name
: str
FDA_get_nonclinical_toxicology_info_by_drug_name
Animal toxicology data
drug_name
: str
FDA_get_carcinogenic_mutagenic_fertility_by_drug_name
Carcinogenicity/mutagenicity/fertility data
drug_name
: str
工具提取的信息参数
FDA_get_boxed_warning_info_by_drug_name
黑框警告（最严重）
drug_name
: str
FDA_get_contraindications_by_drug_name
绝对禁忌症
drug_name
: str
FDA_get_adverse_reactions_by_drug_name
已知不良反应
drug_name
: str
FDA_get_warnings_by_drug_name
警告与注意事项
drug_name
: str
FDA_get_nonclinical_toxicology_info_by_drug_name
动物毒理学数据
drug_name
: str
FDA_get_carcinogenic_mutagenic_fertility_by_drug_name
致癌性/致突变性/生育力数据
drug_name
: str

Workflow

工作流程

  1. Call all 6 FDA tools in parallel (independent queries by drug name)
  2. Parse and structure each response
  3. Prioritize: Boxed Warnings > Contraindications > Warnings > Adverse Reactions
  4. All FDA label data is [T1] evidence (regulatory finding based on human/animal data)
  1. 并行调用所有6个FDA工具（按药物名称独立查询）
  2. 解析并结构化每个工具的返回结果
  3. 优先级: 黑框警告 > 禁忌症 > 警告 > 不良反应
  4. 所有FDA标签数据为[T1]证据（基于人体/动物数据的监管结论）

Decision Logic

决策逻辑

  • Boxed warning present: Flag as CRITICAL safety concern in executive summary
  • No FDA data: Chemical may not be an approved drug; note "Not an FDA-approved drug" and continue with other phases
  • Multiple warnings: Categorize by organ system (hepatic, cardiac, renal, CNS, etc.)
  • Nonclinical toxicology: Grade as [T2] (animal data supporting human risk)
  • 存在黑框警告: 在执行摘要中标记为严重安全风险
  • 无FDA数据: 该化学物质可能未获批为药物；标注“非FDA获批药物”并继续其他阶段分析
  • 多个警告: 按器官系统分类（肝、心脏、肾、中枢神经系统等）
  • 非临床毒理学: 分级为[T2]（支持人体风险的动物数据）

Output Format

输出格式

markdown
undefined
markdown
undefined

Regulatory Safety (FDA) [T1]

监管安全（FDA） [T1]

Boxed Warning

黑框警告

PRESENT - Hepatotoxicity risk with doses >4g/day. Liver failure reported. [T1]
存在 - 日剂量超过4g时存在肝毒性风险，已有肝衰竭病例报告。[T1]

Contraindications

禁忌症

  • Severe hepatic impairment [T1]
  • Known hypersensitivity [T1]
  • 严重肝功能不全 [T1]
  • 已知过敏反应 [T1]

Adverse Reactions (by frequency)

不良反应（按发生频率）

ReactionFrequencySeverity
NauseaCommon (>1%)Mild
HepatotoxicityRare (<0.1%)Severe
.........
反应发生频率严重程度
恶心常见（>1%）轻度
肝毒性罕见（<0.1%）重度
.........

Nonclinical Toxicology [T2]

非临床毒理学 [T2]

  • Carcinogenicity: No carcinogenic potential in 2-year rat/mouse studies
  • Mutagenicity: Negative in Ames assay and in vivo micronucleus test
  • Fertility: No effects on fertility at doses up to 10x human dose

---
  • 致癌性: 2年大鼠/小鼠研究未发现致癌潜力
  • 致突变性: Ames试验及体内微核试验结果为阴性
  • 生育力: 剂量达人体剂量10倍时未发现生育力影响

---

Phase 5: Drug Safety Profile (DrugBank)

阶段5：药物安全档案（DrugBank）

When: Compound is a known drug
Objective: Retrieve curated drug safety data from DrugBank
适用场景: 化合物为已知药物
目标: 从DrugBank获取经过整理的药物安全数据

Tools Used

使用的工具

ToolInformationParameters
drugbank_get_safety_by_drug_name_or_drugbank_id
Toxicity, contraindications
query
: str, 
case_sensitive
: bool, 
exact_match
: bool, 
limit
: int
工具提取的信息参数
drugbank_get_safety_by_drug_name_or_drugbank_id
毒性、禁忌症
query
: str, 
case_sensitive
: bool, 
exact_match
: bool, 
limit
: int

Workflow

工作流程

  1. Call 
    drugbank_get_safety_by_drug_name_or_drugbank_id(query=drug_name, case_sensitive=False, exact_match=False, limit=5)
  2. Parse toxicity information, overdose data, contraindications
  3. Cross-reference with FDA data from Phase 4
  1. 调用
    drugbank_get_safety_by_drug_name_or_drugbank_id(query=drug_name, case_sensitive=False, exact_match=False, limit=5)
  2. 解析毒性信息、过量数据、禁忌症
  3. 与阶段4的FDA数据进行交叉验证

Decision Logic

决策逻辑

  • Toxicity field: Contains LD50 values, overdose symptoms, organ toxicity data
  • DrugBank ID: Note if found for cross-referencing
  • Conflict with FDA: If DrugBank and FDA disagree, note discrepancy and defer to FDA [T1]
  • Not found: Chemical may not be in DrugBank; continue with other phases
  • 毒性字段: 包含LD50值、过量症状、器官毒性数据
  • DrugBank ID: 如果找到，标注以便交叉引用
  • 与FDA数据冲突: 如果DrugBank与FDA数据不一致，标注差异并以FDA[T1]数据为准
  • 未找到数据: 该化学物质可能未收录于DrugBank；继续其他阶段分析

Phase 6: Chemical-Protein Interactions (STITCH)

阶段6：化学-蛋白质相互作用（STITCH）

When: Compound can be identified by name or SMILES
Objective: Map chemical-protein interaction network for off-target assessment
适用场景: 可通过名称或SMILES识别化合物
目标: 绘制化学-蛋白质相互作用网络以评估脱靶效应

Tools Used

使用的工具

ToolFunctionParameters
STITCH_resolve_identifier
Resolve chemical name to STITCH ID
identifier
: str, 
species
: int (9606=human)
STITCH_get_chemical_protein_interactions
Get chemical-protein interactions
identifiers
: list[str], 
species
: int, 
required_score
: int
STITCH_get_interaction_partners
Get interaction network
identifiers
: list[str], 
species
: int, 
limit
: int
工具功能参数
STITCH_resolve_identifier
将化学物质名称解析为STITCH ID
identifier
: str, 
species
: int（9606=人类）
STITCH_get_chemical_protein_interactions
获取化学-蛋白质相互作用
identifiers
: list[str], 
species
: int, 
required_score
: int
STITCH_get_interaction_partners
获取相互作用网络
identifiers
: list[str], 
species
: int, 
limit
: int

Workflow

工作流程

  1. Resolve compound: 
    STITCH_resolve_identifier(identifier=compound_name, species=9606)
  2. Get interactions: 
    STITCH_get_chemical_protein_interactions(identifiers=[stitch_id], species=9606, required_score=700)
  3. Identify off-target proteins (not the intended drug target)
  4. Flag safety-relevant targets: hERG (cardiac), CYP enzymes (metabolism), nuclear receptors (endocrine)
  1. 解析化合物: 
    STITCH_resolve_identifier(identifier=compound_name, species=9606)
  2. 获取相互作用: 
    STITCH_get_chemical_protein_interactions(identifiers=[stitch_id], species=9606, required_score=700)
  3. 识别脱靶蛋白质（非预期药物靶点）
  4. 标记与安全相关的靶点: hERG（心脏）、CYP酶（代谢）、核受体（内分泌）

Decision Logic

决策逻辑

  • High confidence (>900): Well-established interaction [T2]
  • Medium confidence (700-900): Probable interaction [T3]
  • Low confidence (400-700): Possible interaction, needs validation [T4]
  • Safety-relevant targets: Flag interactions with known safety targets
  • No STITCH data: Chemical may be too novel; note and continue
  • 高置信度（>900）: 已充分验证的相互作用 [T2]
  • 中等置信度（700-900）: 可能存在的相互作用 [T3]
  • 低置信度（400-700）: 可能存在的相互作用，需验证 [T4]
  • 安全相关靶点: 标记与已知安全靶点的相互作用
  • 无STITCH数据: 该化学物质可能过于新颖；标注并继续其他分析

Phase 7: Structural Alerts (ChEMBL)

阶段7：结构警报（ChEMBL）

When: ChEMBL molecule ID is available (from Phase 0)
Objective: Check for known toxic substructures
适用场景: 已获取ChEMBL分子ID（来自阶段0）
目标: 检查已知毒性子结构

Tools Used

使用的工具

ToolFunctionParameters
ChEMBL_search_compound_structural_alerts
Find structural alert matches
molecule_chembl_id
: str, 
limit
: int
工具功能参数
ChEMBL_search_compound_structural_alerts
查找结构警报匹配
molecule_chembl_id
: str, 
limit
: int

Workflow

工作流程

  1. If ChEMBL ID available: 
    ChEMBL_search_compound_structural_alerts(molecule_chembl_id=chembl_id, limit=20)
  2. Parse alert types: PAINS (pan-assay interference), Brenk (medicinal chemistry), Glaxo (GSK structural alerts)
  3. Categorize severity: Some alerts are informational, others indicate likely toxicity
  1. 如果有ChEMBL ID: 
    ChEMBL_search_compound_structural_alerts(molecule_chembl_id=chembl_id, limit=20)
  2. 解析警报类型: PAINS（泛试验干扰）、Brenk（药物化学）、Glaxo（GSK结构警报）
  3. 分类严重程度: 部分警报为信息性，部分则提示潜在毒性

Decision Logic

决策逻辑

  • PAINS alerts: May cause false positives in screening; note for medicinal chemistry
  • Brenk alerts: Known problematic substructures; flag if present
  • No alerts: Good sign but not definitive proof of safety
  • No ChEMBL ID: Skip this phase gracefully; note "structural alert analysis not available"
  • PAINS警报: 可能导致筛选假阳性；为药物化学研究标注
  • Brenk警报: 已知问题子结构；如果存在则标记
  • 无警报: 是良好信号，但不能作为安全性的确定性证明
  • 无ChEMBL ID: 优雅跳过该阶段；标注“结构警报分析不可用”

Synthesis: Integrated Risk Assessment (MANDATORY)

合成阶段：综合风险评估（强制要求）

Always the final section. Integrates all evidence into actionable risk classification.
必须为最后一个章节。将所有证据整合为可执行的风险分类。

Risk Classification Matrix

风险分类矩阵

Risk LevelCriteria
CRITICALFDA boxed warning present OR multiple [T1] toxicity findings OR active DILI + active hERG
HIGHFDA warnings present OR [T2] animal toxicity OR multiple active ADMET endpoints
MEDIUMSome [T3] predictions positive OR CTD disease associations OR structural alerts
LOWAll ADMET endpoints negative AND no FDA/DrugBank safety flags AND no CTD concerns
INSUFFICIENT DATAFewer than 3 phases returned data; cannot make confident assessment
风险等级标准
严重存在FDA黑框警告 或 多个[T1]毒性发现 或 活性DILI+活性hERG
高存在FDA警告 或 [T2]动物毒性 或 多个活性ADMET终点
中等部分[T3]预测为阳性 或 CTD疾病关联 或 存在结构警报
低所有ADMET终点为阴性 且 无FDA/DrugBank安全标记 且 无CTD风险
数据不足少于3个阶段返回数据；无法做出可信评估

Synthesis Template

合成模板

markdown
undefined
markdown
undefined

Integrated Risk Assessment

综合风险评估

Overall Risk Classification: [HIGH]

整体风险分类: [高]

Evidence Summary

证据摘要

DimensionFindingEvidence TierConcern
ADMET ToxicityDILI active, hERG active[T3]HIGH
FDA LabelBoxed warning for hepatotoxicity[T1]CRITICAL
CTD Toxicogenomics156 gene interactions, liver neoplasms[T2]HIGH
DrugBankKnown hepatotoxicity at high doses[T2]HIGH
STITCHBinds CYP3A4, hERG[T3]MEDIUM
Structural Alerts2 Brenk alerts[T3]MEDIUM
维度发现证据等级风险
ADMET毒性DILI活性、hERG活性[T3]高
FDA标签存在肝毒性黑框警告[T1]严重
CTD毒理基因组学156个基因相互作用、肝肿瘤关联[T2]高
DrugBank高剂量下存在已知肝毒性[T2]高
STITCH与CYP3A4、hERG结合[T3]中等
结构警报2个Brenk警报[T3]中等

Key Safety Concerns

关键安全风险

  1. Hepatotoxicity [T1]: FDA boxed warning + ADMET-AI DILI prediction + CTD liver disease associations
  2. Cardiac Risk [T3]: ADMET-AI hERG prediction + STITCH hERG interaction
  3. Drug Interactions [T3]: CYP3A4 substrate/inhibitor, potential DDI risk
  1. 肝毒性 [T1]: FDA黑框警告 + ADMET-AI DILI预测 + CTD肝疾病关联
  2. 心脏风险 [T3]: ADMET-AI hERG预测 + STITCH hERG相互作用
  3. 药物相互作用 [T3]: CYP3A4底物/抑制剂，存在潜在药物相互作用风险

Data Gaps

数据缺口

  • No in vivo genotoxicity data available
  • STITCH interaction scores moderate (700-900)
  • No environmental exposure data
  • 无体内遗传毒性数据
  • STITCH相互作用评分中等（700-900）
  • 无环境暴露数据

Recommendations

建议

  1. Avoid doses >4g/day (hepatotoxicity threshold) [T1]
  2. Monitor liver function in chronic use [T1]
  3. Screen for CYP3A4 interactions before co-administration [T3]
  4. Consider cardiac monitoring for at-risk patients [T3]

---
  1. 避免日剂量超过4g（肝毒性阈值） [T1]
  2. 长期使用时监测肝功能 [T1]
  3. 联合用药前筛查CYP3A4相互作用 [T3]
  4. 对高危患者考虑心脏监测 [T3]

---

Mandatory Completeness Checklist

强制完整性检查清单

Before finalizing any report, verify:
  • Phase 0: Compound fully disambiguated (SMILES + CID at minimum)
  • Phase 1: At least 5 toxicity endpoints reported or "prediction unavailable" noted
  • Phase 2: ADMET profile with A/D/M/E sections or "not available" noted
  • Phase 3: CTD queried; gene interactions and disease associations reported or "no data in CTD"
  • Phase 4: FDA labels queried; results or "not an FDA-approved drug" noted
  • Phase 5: DrugBank queried; results or "not found in DrugBank" noted
  • Phase 6: STITCH queried; results or "no STITCH data available" noted
  • Phase 7: Structural alerts checked or "ChEMBL ID not available" noted
  • Synthesis: Risk classification provided with evidence summary
  • Evidence Grading: All findings have [T1]-[T4] annotations
  • Data Gaps: Explicitly listed in synthesis section
在最终确定报告前，需验证:
  • 阶段0: 化合物已完全消歧义（至少包含SMILES + CID）
  • 阶段1: 至少报告5个毒性终点或标注“预测不可用”
  • 阶段2: ADMET分析包含吸收/分布/代谢/排泄章节或标注“不可用”
  • 阶段3: 已查询CTD；报告基因相互作用和疾病关联或标注“CTD无数据”
  • 阶段4: 已查询FDA标签；报告结果或标注“非FDA获批药物”
  • 阶段5: 已查询DrugBank；报告结果或标注“未收录于DrugBank”
  • 阶段6: 已查询STITCH；报告结果或标注“无STITCH数据”
  • 阶段7: 已检查结构警报或标注“ChEMBL ID不可用”
  • 合成阶段: 提供风险分类及证据摘要
  • 证据分级: 所有发现均有[T1]-[T4]标注
  • 数据缺口: 在合成阶段明确列出

Tool Parameter Reference

工具参数参考

Critical Parameter Notes (verified from source code):
ToolParameter NameTypeNotes
All ADMETAI tools
smiles
list[str]
Always a list, even for single compound
All CTD tools
input_terms
str
Chemical name, MeSH name, CAS RN, or MeSH ID
All FDA tools
drug_name
str
Brand or generic drug name
drugbank_get_safety_*
query
, 
case_sensitive
, 
exact_match
, 
limit
str, bool, bool, intAll 4 required
STITCH_resolve_identifier
identifier
, 
species
str, intspecies=9606 for human
STITCH_get_chemical_protein_interactions
identifiers
, 
species
, 
required_score
list[str], int, intrequired_score=400 default
PubChem_get_CID_by_compound_name
name
str
Compound name (not SMILES)
PubChem_get_compound_properties_by_CID
cid
int
Numeric CID
ChEMBL_search_compound_structural_alerts
molecule_chembl_id
str
ChEMBL ID (e.g., "CHEMBL112")
关键参数说明（已从源代码验证）:
工具参数名称类型说明
所有ADMETAI工具
smiles
list[str]
始终为列表，即使单个化合物
所有CTD工具
input_terms
str
化学物质名称、MeSH名称、CAS号或MeSH ID
所有FDA工具
drug_name
str
商品名或通用药物名称
drugbank_get_safety_*
query
, 
case_sensitive
, 
exact_match
, 
limit
str, bool, bool, int4个参数均为必填
STITCH_resolve_identifier
identifier
, 
species
str, intspecies=9606代表人类
STITCH_get_chemical_protein_interactions
identifiers
, 
species
, 
required_score
list[str], int, intrequired_score默认值为400
PubChem_get_CID_by_compound_name
name
str
化合物名称（非SMILES）
PubChem_get_compound_properties_by_CID
cid
int
数字CID
ChEMBL_search_compound_structural_alerts
molecule_chembl_id
str
ChEMBL ID（如“CHEMBL112”）

Response Format Notes

返回格式说明

  • ADMET-AI: Returns 
    {status: "success", data: {...}}
    with prediction values
  • CTD: Returns list of interaction/association objects
  • FDA: Returns 
    {status, data}
    with label text
  • DrugBank: Returns 
    {data: [...]}
    with drug records
  • STITCH: Returns list of interaction objects with scores
  • PubChem CID lookup: Returns 
    {IdentifierList: {CID: [...]}}
    (may or may not have 
    data
    wrapper)
  • PubChem properties: Returns dict with 
    CID
    , 
    MolecularWeight
    , 
    ConnectivitySMILES
    , 
    IUPACName
  • ADMET-AI: 返回
    {status: "success", data: {...}}
    包含预测值
  • CTD: 返回相互作用/关联对象列表
  • FDA: 返回
    {status, data}
    包含标签文本
  • DrugBank: 返回
    {data: [...]}
    包含药物记录
  • STITCH: 返回带评分的相互作用对象列表
  • PubChem CID查询: 返回
    {IdentifierList: {CID: [...]}}
    （可能包含或不包含
    data
    包装）
  • PubChem属性查询: 返回包含
    CID
    、
    MolecularWeight
    、
    ConnectivitySMILES
    、
    IUPACName
    的字典

Fallback Strategies

回退策略

Compound Resolution

化合物解析

  • Primary: PubChem by name -> CID -> properties -> SMILES
  • Fallback 1: ChEMBL search by name -> molecule -> SMILES
  • Fallback 2: If SMILES provided directly, skip name resolution
  • 主策略: PubChem按名称查询→CID→属性→SMILES
  • 回退1: ChEMBL按名称查询→分子→SMILES
  • 回退2: 如果直接提供SMILES，跳过名称解析

Toxicity Prediction

毒性预测

  • Primary: All 9 ADMET-AI endpoints
  • Fallback: If ADMET-AI fails for a compound, note "prediction failed" and continue with database evidence
  • Note: ADMET-AI may fail for very large or unusual SMILES
  • 主策略: 所有9个ADMET-AI终点
  • 回退: 如果ADMET-AI对某化合物调用失败，标注“预测失败”并继续使用数据库证据
  • 注意: ADMET-AI可能对超大或特殊SMILES调用失败

Regulatory Data

监管数据

  • Primary: FDA labels by drug name
  • Fallback: If FDA returns no data, try alternative drug names (brand vs generic)
  • Note: Non-drug chemicals (pesticides, industrial) will not have FDA labels
  • 主策略: 按药物名称查询FDA标签
  • 回退: 如果FDA未返回数据，尝试替代药物名称（商品名vs通用名）
  • 注意: 非药物化学物质（农药、工业化学品）无FDA标签

CTD Data

CTD数据

  • Primary: Search by common chemical name
  • Fallback: Try MeSH name if common name fails
  • Note: Novel compounds may not be in CTD
  • 主策略: 按常用化学物质名称查询
  • 回退: 如果常用名称失败，尝试MeSH名称
  • 注意: 新型化合物可能未收录于CTD

Common Use Patterns

常见使用模式

Pattern 1: Novel Compound Assessment

模式1：新型化合物评估

Input: SMILES string for new molecule
Workflow: Phase 0 (SMILES->CID) -> Phase 1 (toxicity) -> Phase 2 (ADMET) -> Phase 7 (structural alerts) -> Synthesis
Output: Predictive safety profile for novel compound
输入: 新型分子的SMILES字符串
工作流程: 阶段0（SMILES→CID）→阶段1（毒性预测）→阶段2（ADMET分析）→阶段7（结构警报）→合成阶段
输出: 新型化合物的预测安全档案

Pattern 2: Approved Drug Safety Review

模式2：获批药物安全审查

Input: Drug name (e.g., "Acetaminophen")
Workflow: All phases (0-7 + Synthesis)
Output: Complete safety dossier with regulatory + predictive + database evidence
输入: 药物名称（如“Acetaminophen”）
工作流程: 所有阶段（0-7 + 合成阶段）
输出: 包含监管、预测、数据库证据的完整安全档案

Pattern 3: Environmental Chemical Risk

模式3：环境化学风险评估

Input: Chemical name (e.g., "Bisphenol A")
Workflow: Phase 0 -> Phase 1 -> Phase 2 -> Phase 3 (CTD, key for env chemicals) -> Phase 6 -> Synthesis
Output: Environmental health risk assessment focused on gene-disease associations
输入: 化学物质名称（如“Bisphenol A”）
工作流程: 阶段0→阶段1→阶段2→阶段3（CTD，对环境化学品至关重要）→阶段6→合成阶段
输出: 聚焦基因-疾病关联的环境健康风险评估

Pattern 4: Batch Toxicity Screening

模式4：批量毒性筛选

Input: Multiple SMILES strings
Workflow: Phase 0 -> Phase 1 (batch) -> Phase 2 (batch) -> Comparative table -> Synthesis
Output: Comparative toxicity table ranking compounds by safety
输入: 多个SMILES字符串
工作流程: 阶段0→阶段1（批量）→阶段2（批量）→对比表格→合成阶段
输出: 按安全性排序的化合物毒性对比表格

Pattern 5: Toxicogenomic Deep-Dive

模式5：毒理基因组学深度分析

Input: Chemical name + specific gene or disease interest
Workflow: Phase 0 -> Phase 3 (CTD expanded) -> Literature search -> Synthesis
Output: Detailed chemical-gene-disease mechanistic analysis
输入: 化学物质名称 + 特定基因或疾病研究需求
工作流程: 阶段0→阶段3（扩展CTD分析）→文献检索→合成阶段
输出: 详细的化学-基因-疾病机制分析

Output Report Structure

输出报告结构

All analyses generate a structured markdown report with progressive sections:
markdown
undefined
所有分析生成结构化markdown报告，包含以下递进章节:
markdown
undefined

Chemical Safety & Toxicology Report: [Compound Name]

化学安全与毒理学报告: [化合物名称]

Generated: YYYY-MM-DD HH:MM Compound: [Name] | SMILES: [SMILES] | CID: [CID]
生成时间: YYYY-MM-DD HH:MM 化合物: [名称] | SMILES: [SMILES] | CID: [CID]

Executive Summary

执行摘要

[2-3 sentence overview with risk classification and key findings, all graded]
[2-3句话概述，包含风险分类和关键发现，均需分级]

1. Compound Identity

1. 化合物身份信息

[Phase 0 results - disambiguation table]
[阶段0结果 - 消歧表格]

2. Predictive Toxicology

2. 预测毒理学

[Phase 1 results - ADMET-AI toxicity endpoints]
[阶段1结果 - ADMET-AI毒性终点]

3. ADMET Profile

3. ADMET分析结果

[Phase 2 results - absorption, distribution, metabolism, excretion]
[阶段2结果 - 吸收、分布、代谢、排泄]

4. Toxicogenomics

4. 毒理基因组学

[Phase 3 results - CTD chemical-gene-disease relationships]
[阶段3结果 - CTD化学-基因-疾病关系]

5. Regulatory Safety

5. 监管安全

[Phase 4 results - FDA label information]
[阶段4结果 - FDA标签信息]

6. Drug Safety Profile

6. 药物安全档案

[Phase 5 results - DrugBank data]
[阶段5结果 - DrugBank数据]

7. Chemical-Protein Interactions

7. 化学-蛋白质相互作用

[Phase 6 results - STITCH network]
[阶段6结果 - STITCH网络]

8. Structural Alerts

8. 结构警报

[Phase 7 results - ChEMBL alerts]
[阶段7结果 - ChEMBL警报]

9. Integrated Risk Assessment

9. 综合风险评估

[Synthesis - risk classification, evidence summary, data gaps, recommendations]
[合成阶段 - 风险分类、证据摘要、数据缺口、建议]

Appendix: Methods and Data Sources

附录: 方法与数据来源

[Tool versions, databases queried, date of access]

---
[工具版本、查询的数据库、访问日期]

---

Limitations & Known Issues

局限性与已知问题

Tool-Specific

工具特定

  • ADMET-AI: Predictions are computational [T3]; should not replace experimental testing
  • CTD: Curated but may lag behind latest literature by 6-12 months
  • FDA: Only covers FDA-approved drugs; not applicable to environmental chemicals or supplements
  • DrugBank: Primarily drugs; limited coverage of industrial chemicals
  • STITCH: Score thresholds affect sensitivity; lower scores increase false positives
  • ChEMBL: Structural alerts require ChEMBL ID; not all compounds have one
  • ADMET-AI: 预测为计算得出的[T3]证据；不能替代实验测试
  • CTD: 经过整理但可能滞后于最新文献6-12个月
  • FDA: 仅覆盖FDA获批药物；不适用于环境化学品或补充剂
  • DrugBank: 主要收录药物；工业化学品覆盖有限
  • STITCH: 评分阈值影响灵敏度；低评分会增加假阳性
  • ChEMBL: 结构警报需要ChEMBL ID；并非所有化合物都有

Analysis

分析相关

  • Novel compounds: May only have ADMET-AI predictions (no database evidence)
  • Environmental chemicals: FDA/DrugBank phases will be empty; rely on CTD and ADMET-AI
  • Batch mode: ADMET-AI can handle batches; other tools require individual queries
  • Species specificity: Most data is human-centric; animal data noted where applicable
  • 新型化合物: 可能仅能获取ADMET-AI预测结果（无数据库证据）
  • 环境化学品: FDA/DrugBank阶段将无数据；依赖CTD和ADMET-AI
  • 批量模式: ADMET-AI支持批量处理；其他工具需单独查询
  • 物种特异性: 大多数数据以人类为中心；动物数据会单独标注

Technical

技术相关

  • SMILES validity: Invalid SMILES will cause ADMET-AI failures
  • Name ambiguity: Chemical names can be ambiguous; always verify with CID
  • Rate limits: Some FDA endpoints may rate-limit for rapid queries
  • SMILES有效性: 无效SMILES会导致ADMET-AI调用失败
  • 名称歧义: 化学物质名称可能存在歧义；始终用CID验证
  • 速率限制: 部分FDA端点可能对快速查询进行速率限制

Summary

总结

Chemical Safety & Toxicology Assessment Skill provides comprehensive safety evaluation by integrating:
  1. Predictive toxicology (ADMET-AI) - 9 tools covering toxicity, ADMET, physicochemical properties
  2. Toxicogenomics (CTD) - Chemical-gene-disease relationship mapping
  3. Regulatory safety (FDA) - 6 tools for label-based safety extraction
  4. Drug safety (DrugBank) - Curated toxicity and contraindication data
  5. Chemical interactions (STITCH) - Chemical-protein interaction networks
  6. Structural alerts (ChEMBL) - Known toxic substructure detection
Outputs: Structured markdown report with risk classification, evidence grading, and actionable recommendations
Best for: Drug safety assessment, chemical hazard profiling, environmental toxicology, ADMET characterization, toxicogenomic analysis
Total tools integrated: 25+ tools across 6 databases
化学安全与毒理学评估技能通过整合以下内容提供全面的安全评估:
  1. 预测毒理学（ADMET-AI）- 9个工具覆盖毒性、ADMET、理化特性
  2. 毒理基因组学（CTD）- 化学-基因-疾病关系映射
  3. 监管安全（FDA）- 6个工具提取标签安全信息
  4. 药物安全（DrugBank）- 经过整理的毒性和禁忌症数据
  5. 化学相互作用（STITCH）- 化学-蛋白质相互作用网络
  6. 结构警报（ChEMBL）- 已知毒性子结构检测
输出: 带有风险分类、证据分级和可执行建议的结构化markdown报告
最佳适用场景: 药物安全评估、化学危害分析、环境毒理学、ADMET特征分析、毒理基因组学分析
整合的工具总数: 6个数据库中的25+个工具