tooluniverse

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

ToolUniverse General Usage Strategies

ToolUniverse通用使用策略

Master strategies for using ToolUniverse's 10000+ scientific tools effectively. These principles apply regardless of how you access ToolUniverse (MCP server, SDK, or direct tool calls).

掌握有效使用ToolUniverse的10000+科学工具的策略。这些原则适用于所有ToolUniverse的访问方式（MCP服务器、SDK或直接工具调用）。

Core Philosophy

核心理念

ToolUniverse has MANY tools - the challenge is discovering and using them effectively:

Search widely - Don't assume you know all relevant tools; always search for more
Query multiple databases - Cross-reference data across sources
Multi-hop persistence - Many answers require 3-5 tool calls in sequence
Never give up easily - If one tool fails, try alternatives
Comprehensive reports - Use all available data; detail is valuable
English-first queries - Always use English terms in tool calls, even if the user writes in another language

ToolUniverse拥有大量工具——挑战在于如何高效发现和使用它们：

广泛搜索 - 不要假设你了解所有相关工具；始终要搜索更多工具
查询多个数据库 - 跨来源交叉引用数据
多跳持续查询 - 许多答案需要连续3-5次工具调用
不要轻易放弃 - 如果一个工具失败，尝试替代工具
生成全面报告 - 利用所有可用数据；细节很有价值
优先使用英文查询 - 即使用户使用其他语言提问，工具调用时也始终使用英文术语

Step 0: Clarify the Request Before Researching

步骤0：研究前先明确需求

CRITICAL: Before starting any research, ensure you understand what the user actually needs. Wasted tool calls on the wrong entity or scope are expensive.

至关重要：在开始任何研究之前，确保你理解用户的实际需求。针对错误实体或范围的无效工具调用成本很高。

When to Ask Clarifying Questions

何时需要询问澄清问题

Signal	Example	What to Clarify
Vague entity	"Research cancer"	Which cancer type? Which aspect (treatment, genetics, epidemiology)?
Ambiguous name	"Tell me about JAK"	JAK1/2/3? The gene family? A specific inhibitor?
Unclear scope	"Look into metformin"	Drug profile? Repurposing? Safety? Mechanism?
Missing context	"What targets this?"	Which compound/disease/pathway?
Multiple interpretations	"ACE"	ACE the gene? ACE inhibitors? ACE2?

信号	示例	需要澄清的内容
模糊实体	"研究癌症"	哪种癌症类型？哪个方面（治疗、遗传学、流行病学）？
名称歧义	"告诉我关于JAK的信息"	JAK1/2/3？基因家族？特定抑制剂？
范围不明确	"研究二甲双胍"	药物概况？重定位？安全性？作用机制？
缺少上下文	"什么靶向这个？"	哪种化合物/疾病/通路？
多种解读	"ACE"	ACE基因？ACE抑制剂？ACE2？

When NOT to Ask

何时无需询问

Proceed directly when the request is specific enough:

"What is the structure of EGFR kinase domain?" - Clear entity + clear data type
"Find all drugs targeting BRAF V600E" - Specific variant + clear task
"Research Alzheimer's disease comprehensively" - Broad but unambiguous

当需求足够明确时直接进行：

"EGFR激酶域的结构是什么？" - 实体明确 + 数据类型明确
"找到所有靶向BRAF V600E的药物" - 特定变异体 + 任务明确
"全面研究阿尔茨海默病" - 范围宽泛但无歧义

Clarification Checklist

澄清检查清单

Before starting research, confirm you know:

Entity - Exactly which gene/protein/drug/disease?
Species - Human unless stated otherwise
Scope - Comprehensive profile or specific aspect?
Output - Report, data table, quick answer, or comparison?

If any of these are unclear, ask the user one concise question covering all ambiguities rather than asking multiple rounds of questions.

开始研究前，确认你了解：

实体 - 具体是哪个基因/蛋白质/药物/疾病？
物种 - 除非特别说明，默认为人类
范围 - 全面概况还是特定方面？
输出形式 - 报告、数据表、快速答案还是对比分析？

如果有任何一项不明确，向用户提出一个简洁的问题涵盖所有歧义点，而非多轮提问。

Strategy 1: Exhaustive Tool Discovery

策略1：全面的工具发现

CRITICAL: ToolUniverse has 10000+ tools. Before any research task, search for ALL relevant tools.

至关重要：ToolUniverse拥有10000+工具。在任何研究任务开始前，搜索所有相关工具。

Tool Discovery Methods

工具发现方法

Use the tool finder tools to discover what's available:

Method	Tool	Best For
Keyword	`Tool_Finder_Keyword`	Fast search by terms
LLM-based	`Tool_Finder_LLM`	Intelligent matching by description
Embedding	`Tool_Finder`	Semantic similarity search

使用工具查找工具来发现可用资源：

方法	工具	最佳适用场景
关键词搜索	`Tool_Finder_Keyword`	通过术语快速搜索
基于LLM的搜索	`Tool_Finder_LLM`	通过描述智能匹配
嵌入搜索	`Tool_Finder`	语义相似度搜索

Discovery Best Practices

发现最佳实践

Practice	Why	Example
Search with multiple terms	Same data from different angles	"protein expression", "gene expression", "tissue expression"
Search by database name	Find all tools for a source	"UniProt", "ChEMBL", "OpenTargets"
Search by data type	Comprehensive coverage	"variant", "mutation", "SNP", "polymorphism"
Search by use case	Task-oriented discovery	"druggability", "target validation"

实践	原因	示例
使用多个术语搜索	从不同角度获取相同数据	"protein expression", "gene expression", "tissue expression"
按数据库名称搜索	查找针对某一来源的所有工具	"UniProt", "ChEMBL", "OpenTargets"
按数据类型搜索	全面覆盖	"variant", "mutation", "SNP", "polymorphism"
按使用场景搜索	面向任务的发现	"druggability", "target validation"

Minimum Discovery Queries

最低要求的发现查询

Before any research task, run at least these searches:

Main topic query:
```
[your topic]
```
Related terms:
```
[synonym1]
```
,
```
[synonym2]
```
Database-specific:
```
[relevant database names]
```
Data type specific:
```
[required data types]
```

Example for target research:

"protein information"
"gene expression"
"drug target"
"UniProt", "OpenTargets"
"protein interaction"
"variant mutation"

在任何研究任务开始前，至少运行以下搜索：

主主题查询：
```
[你的主题]
```
相关术语：
```
[同义词1]
```
,
```
[同义词2]
```
特定数据库：
```
[相关数据库名称]
```
特定数据类型：
```
[所需数据类型]
```

靶点研究示例：

"protein information"
"gene expression"
"drug target"
"UniProt", "OpenTargets"
"protein interaction"
"variant mutation"

Strategy 2: Multi-Hop Tool Chains

策略2：多跳工具链

CRITICAL: Most scientific questions require multiple tool calls. A single tool rarely gives the complete answer.

至关重要：大多数科学问题需要多次工具调用。单一工具很少能给出完整答案。

Why Multi-Hop Matters

多跳的重要性

Question Type	Single Tool Answer	Multi-Hop Answer
"Tell me about EGFR"	Basic protein info	Full profile with structure, expression, drugs, variants, literature
"What drugs target TP53?"	List of drug names	Drug details, mechanisms, clinical trials, bioactivity data
"Research Alzheimer's"	Disease definition	Genes, pathways, drugs, trials, phenotypes, GWAS, literature

问题类型	单一工具答案	多跳答案
"告诉我关于EGFR的信息"	基础蛋白质信息	包含结构、表达、药物、变异体、文献的完整概况
"哪些药物靶向TP53？"	药物名称列表	药物详情、作用机制、临床试验、生物活性数据
"研究阿尔茨海默病"	疾病定义	基因、通路、药物、试验、表型、GWAS、文献

Common Multi-Hop Patterns

常见多跳模式

Pattern A: ID Resolution Chain

模式A：ID解析链

Name → ID → Data → Related Data

Example: Gene name to complete profile
1. gene_name → Ensembl ID
2. Ensembl ID → UniProt accession  
3. UniProt accession → Protein entry
4. UniProt accession → Domains
5. UniProt accession → Structure

名称 → ID → 数据 → 相关数据

示例：从基因名称到完整概况
1. gene_name → Ensembl ID
2. Ensembl ID → UniProt accession  
3. UniProt accession → Protein entry
4. UniProt accession → Domains
5. UniProt accession → Structure

Pattern B: Cross-Database Enrichment

模式B：跨数据库富集

Primary Data → Cross-reference → Enriched View

Example: Drug compound enrichment
1. drug_name → PubChem CID
2. drug_name → ChEMBL ID
3. CID → properties
4. ChEMBL ID → bioactivity
5. ChEMBL ID → targets
6. SMILES → ADMET predictions

原始数据 → 交叉引用 → 丰富视图

示例：药物化合物富集
1. drug_name → PubChem CID
2. drug_name → ChEMBL ID
3. CID → 性质
4. ChEMBL ID → 生物活性
5. ChEMBL ID → 靶点
6. SMILES → ADMET预测

Pattern C: Network Expansion

模式C：网络扩展

Seed Entity → Connected Entities → Entity Details

Example: Target interaction network
1. gene → protein interactions
2. For each interactor → gene info
3. Interactor → disease associations

种子实体 → 关联实体 → 实体详情

示例：靶点互作网络
1. gene → 蛋白质互作
2. 针对每个互作蛋白 → 基因信息
3. 互作蛋白 → 疾病关联

Pattern D: Literature + Data Integration

模式D：文献+数据整合

Database Annotations → Literature Evidence → Synthesis

Example: Disease mechanism research
1. disease → associated genes
2. disease → phenotypes
3. disease → drugs
4. disease → literature
5. key papers → citations

数据库注释 → 文献证据 → 综合分析

示例：疾病机制研究
1. disease → 关联基因
2. disease → 表型
3. disease → 药物
4. disease → 文献
5. 关键论文 → 引用文献

Multi-Hop Persistence Rules

多跳持续规则

Don't stop at first result - One tool gives partial data; keep going
Follow cross-references - Use IDs from one tool to query others
Chain until complete - 5-10 tool calls for comprehensive answers is normal
Track all IDs - Save every identifier for potential future use

不要停留在第一个结果 - 单一工具只能提供部分数据；继续深入
跟踪交叉引用 - 使用一个工具的ID查询其他工具
持续直到完整 - 全面回答通常需要5-10次工具调用
记录所有ID - 保存每个标识符以备未来使用

Strategy 3: Query Multiple Databases for Same Data

策略3：针对同一数据查询多个数据库

CRITICAL: Different databases have different coverage. Query ALL relevant sources.

至关重要：不同数据库的覆盖范围不同。查询所有相关来源。

Database Redundancy Principle

数据库冗余原则

For any data type, query multiple sources:

Data Type	Primary	Secondary	Tertiary
Protein info	UniProt	Proteins API	NCBI Protein
Gene expression	GTEx	Human Protein Atlas	ArrayExpress
Drug targets	ChEMBL	DGIdb	OpenTargets
Variants	gnomAD	ClinVar	OpenTargets
Literature	PubMed	Europe PMC	OpenAlex
Pathways	Reactome	KEGG	WikiPathways
Structures	RCSB PDB	PDBe	AlphaFold
Disease associations	OpenTargets	ClinVar	GWAS Catalog

对于任何数据类型，查询多个来源：

数据类型	主要来源	次要来源	三级来源
蛋白质信息	UniProt	Proteins API	NCBI Protein
基因表达	GTEx	Human Protein Atlas	ArrayExpress
药物靶点	ChEMBL	DGIdb	OpenTargets
变异体	gnomAD	ClinVar	OpenTargets
文献	PubMed	Europe PMC	OpenAlex
通路	Reactome	KEGG	WikiPathways
结构	RCSB PDB	PDBe	AlphaFold
疾病关联	OpenTargets	ClinVar	GWAS Catalog

Merge Results Strategy

结果合并策略

When querying multiple databases:

Collect all results - Don't stop at first success
Note data source - Track where each datum came from
Handle conflicts - Document when sources disagree
Prefer curated - Weight RefSeq over GenBank, UniProt over predictions

查询多个数据库时：

收集所有结果 - 不要在第一次成功后停止
记录数据源 - 跟踪每条数据的来源
处理冲突 - 记录来源之间的不一致
优先选择 curated数据 - 优先选择RefSeq而非GenBank，UniProt而非预测数据

Strategy 3.1: Abstract Search vs Full-Text Search (Literature)

策略3.1：摘要搜索 vs 全文搜索（文献）

CRITICAL: Many biomedical “needle” terms (rsIDs like

rs58542926

, reagent catalog numbers, supplementary-table IDs) never appear in titles/abstracts. If you only search abstracts, you will miss papers even when they are open access.

至关重要：许多生物医学“精准”术语（如rsID

rs58542926

、试剂目录号、补充表格ID）从未出现在标题/摘要中。如果仅搜索摘要，即使论文是开放获取的，你也会错过相关内容。

Quick rule

快速规则

If your keywords look like body-only terms (rsIDs, figure/table references, “Supplementary Table”), use full-text-aware tools first.

如果你的关键词看起来是仅正文出现的术语（rsID、图/表引用、“补充表格”），首先使用支持全文搜索的工具。

Tools that can match full text (indexed or retrieved)

可匹配全文的工具（索引或检索）

Goal	Tools	Notes
Indexed full-text search (biomed OA)	`PMC_search_papers`	NCBI “pmc” database indexes full text; good for rsIDs.
Indexed full-text search (Europe PMC subset)	`EuropePMC_search_articles` with `require_has_ft=true` + `fulltext_terms=[...]`	Uses Europe PMC `HAS_FT:Y` + `BODY:\"...\"` fielded queries; works only when Europe PMC has indexed full text.
Best-effort full-text retrieval + keyword snippets	`EuropePMC_get_fulltext_snippets`	Fetches full text (XML → HTML fallbacks) and returns bounded snippets with `retrieval_trace` .
OA aggregation + (sometimes) full-text search	`CORE_search_papers`	Coverage varies; a paper may not exist in CORE even if OA elsewhere.
Download-and-scan fallback	`CORE_get_fulltext_snippets`	Local PDF scan for body-only terms when index-based search misses; can fail if the “PDF” URL returns HTML/403 (check trace/content-type).
Partial full-text indexing (not guaranteed)	`openalex_search_works` / `openalex_literature_search` with `require_has_fulltext` / `fulltext_terms`	Only matches works where OpenAlex has indexed full text; can miss PMC-hosted full text. Use as a secondary signal.

目标	工具	说明
索引全文搜索（生物医学OA）	`PMC_search_papers`	NCBI“pmc”数据库索引全文；适合rsID搜索。
索引全文搜索（Europe PMC子集）	`EuropePMC_search_articles` 搭配 `require_has_ft=true` + `fulltext_terms=[...]`	使用Europe PMC的 `HAS_FT:Y` + `BODY:\"...\"` 字段查询；仅当Europe PMC已索引全文时有效。
最佳尝试全文检索+关键词片段	`EuropePMC_get_fulltext_snippets`	获取全文（XML→HTML备选）并返回带 `retrieval_trace` 的限定片段。
OA聚合+（有时）全文搜索	`CORE_search_papers`	覆盖范围不一；即使论文是OA的，也可能未收录在CORE中。
下载并扫描备选方案	`CORE_get_fulltext_snippets`	当基于索引的搜索失败时，本地PDF扫描正文专属术语；如果“PDF”URL返回HTML/403则可能失败（检查trace/content-type）。
部分全文索引（不保证）	`openalex_search_works` / `openalex_literature_search` 搭配 `require_has_fulltext` / `fulltext_terms`	仅匹配OpenAlex已索引全文的文献；可能错过PMC托管的全文。作为次要信号使用。

Recommended flow for body-only keywords

正文专属关键词的推荐流程

Try

PMC_search_papers

and

EuropePMC_search_articles

(with

require_has_ft

fulltext_terms

If you have a PMCID/PMID, use
```
EuropePMC_get_fulltext_snippets
```
to confirm the term is in the paper.
If you only have a PDF URL, use
```
CORE_get_fulltext_snippets
```
as a last resort, and treat HTTP
```
200
```
as “request succeeded”, not “PDF succeeded” (validate
```
content_type
```
).

尝试

PMC_search_papers

和

EuropePMC_search_articles

（搭配

require_has_ft

fulltext_terms

）。

如果有PMCID/PMID，使用
```
EuropePMC_get_fulltext_snippets
```
确认术语是否在论文中。
如果只有PDF URL，最后尝试
```
CORE_get_fulltext_snippets
```
，将HTTP
```
200
```
视为“请求成功”而非“PDF获取成功”（验证
```
content_type
```
）。

Strategy 4: Disambiguation First

策略4：先消除歧义

CRITICAL: Before any research, resolve entity identity to avoid wrong data and missed results.

至关重要：在任何研究之前，先解析实体身份，避免错误数据和遗漏结果。

Why Disambiguation Matters

消除歧义的重要性

Problem	Example	Consequence
Naming collision	"JAK" = Janus kinase OR "just another kinase"	Wrong papers retrieved
Multiple IDs	Gene has symbol, Ensembl, Entrez, UniProt IDs	Miss data in some databases
Salt forms	Metformin vs metformin HCl (different CIDs)	Incomplete compound data
Species ambiguity	BRCA1 in human vs mouse	Wrong expression/function data

问题	示例	后果
命名冲突	"JAK" = Janus激酶 OR "just another kinase"	检索到错误论文
多个ID	基因有符号、Ensembl、Entrez、UniProt ID	遗漏部分数据库的数据
盐形式	二甲双胍 vs 盐酸二甲双胍（不同CID）	化合物数据不完整
物种歧义	人类 vs 小鼠的BRCA1	表达/功能数据错误

Disambiguation Workflow

歧义消除工作流

Step 1: Establish Canonical IDs
    gene_name → UniProt, Ensembl, NCBI Gene, ChEMBL target
    compound_name → PubChem CID, ChEMBL ID, SMILES
    disease_name → EFO ID, ICD-10, UMLS CUI

Step 2: Gather Synonyms
    All aliases, alternative names, historical names
    
Step 3: Detect Naming Collisions
    Search "[TERM]"[Title] → check if results are on-topic
    Build negative filters: NOT [collision_term]
    
Step 4: Species Confirmation
    Verify organism is correct (default: Homo sapiens)

步骤1：建立标准ID
    gene_name → UniProt, Ensembl, NCBI Gene, ChEMBL target
    compound_name → PubChem CID, ChEMBL ID, SMILES
    disease_name → EFO ID, ICD-10, UMLS CUI

步骤2：收集同义词
    所有别名、替代名称、历史名称
    
步骤3：检测命名冲突
    搜索"[TERM]"[Title] → 检查结果是否相关
    构建负面过滤器：NOT [冲突术语]
    
步骤4：物种确认
    验证生物是否正确（默认：Homo sapiens）

ID Types by Entity

按实体分类的ID类型

Genes/Proteins:

Gene Symbol (EGFR, TP53)
UniProt accession (P00533)
Ensembl ID (ENSG00000146648)
NCBI Gene ID (1956)
ChEMBL Target ID (CHEMBL203)

Compounds:

PubChem CID (2244)
ChEMBL ID (CHEMBL25)
SMILES string
InChI/InChIKey

Diseases:

EFO ID (EFO_0000249)
ICD-10 code (G30)
UMLS CUI (C0002395)
SNOMED CT

基因/蛋白质：

基因符号（EGFR, TP53）
UniProt accession（P00533）
Ensembl ID（ENSG00000146648）
NCBI Gene ID（1956）
ChEMBL Target ID（CHEMBL203）

化合物：

PubChem CID（2244）
ChEMBL ID（CHEMBL25）
SMILES字符串
InChI/InChIKey

疾病：

EFO ID（EFO_0000249）
ICD-10代码（G30）
UMLS CUI（C0002395）
SNOMED CT

Strategy 5: Never Give Up on Search

策略5：搜索永不放弃

CRITICAL: When a tool fails or returns empty, don't give up. Try alternatives.

至关重要：当工具失败或返回空结果时，不要放弃。尝试替代方案。

Failure Handling Protocol

故障处理流程

Attempt 1: Primary tool
    ↓ fails
Wait briefly, then retry
    ↓ fails
Try fallback tool #1
    ↓ fails
Try fallback tool #2
    ↓ fails
Document as "unavailable" with reason

尝试1：主工具
    ↓ 失败
短暂等待后重试
    ↓ 失败
尝试备选工具#1
    ↓ 失败
尝试备选工具#2
    ↓ 失败
记录为“不可用”并说明原因

Common Fallback Chains

常见备选链

Primary Tool	Fallback Options
PubMed citations	EuropePMC citations → OpenAlex citations
GTEx expression	Human Protein Atlas expression
PubChem compound lookup	ChEMBL search → SMILES-based lookup
ChEMBL bioactivity	PubChem bioactivity summary
DailyMed drug labels	PubChem drug label info
UniProt protein entry	Proteins API

主工具	备选选项
PubMed引用	EuropePMC引用 → OpenAlex引用
GTEx表达	Human Protein Atlas表达
PubChem化合物查询	ChEMBL搜索 → 基于SMILES的查询
ChEMBL生物活性	PubChem生物活性摘要
DailyMed药物标签	PubChem药物标签信息
UniProt蛋白质条目	Proteins API

Alternative Search Strategies

替代搜索策略

If keyword search fails:

Try synonyms and aliases
Use broader/narrower terms
Try different databases

If database is empty:

Query related databases
Use literature to find mentions
Check if entity exists under different name

If API rate-limited:

Wait and retry
Try same query on different database
Use cached results if available

如果关键词搜索失败：

尝试同义词和别名
使用更宽泛/更具体的术语
尝试不同数据库

如果数据库为空：

查询相关数据库
使用文献查找提及内容
检查实体是否以其他名称存在

如果API速率受限：

等待并重试
在不同数据库尝试相同查询
如果可用，使用缓存结果

Strategy 6: Generate Comprehensive Reports

策略6：生成全面报告

CRITICAL: With access to many tools, reports should be detailed and thorough.

至关重要：借助众多工具的访问权限，报告应详细且全面。

Report-First Approach

以报告为导向的方法

Create report structure FIRST - Define all sections before gathering data
Progressively update - Fill sections as data is gathered
Show findings, not process - Report results, not search methodology

先创建报告结构 - 在收集数据前定义所有章节
逐步更新 - 收集数据时填充章节内容
展示发现，而非过程 - 报告结果，而非搜索方法

Citation Requirements

引用要求

Every fact must have a source:

undefined

每个事实都必须有来源：

undefined

Protein Function

蛋白质功能

EGFR is a receptor tyrosine kinase that regulates cell growth. Source: UniProt (P00533)

EGFR是一种调节细胞生长的受体酪氨酸激酶。 来源：UniProt (P00533)

Expression Profile

表达谱

Tissue	TPM	Source
Skin	156.3	GTEx
Lung	98.4	GTEx

undefined

组织	TPM	来源
皮肤	156.3	GTEx
肺	98.4	GTEx

undefined

Evidence Grading

证据分级

Grade claims by evidence strength:

Tier	Symbol	Description	Example
T1	★★★	Mechanistic with direct evidence	CRISPR KO study
T2	★★☆	Functional study	siRNA knockdown
T3	★☆☆	Association/screen hit	GWAS, high-throughput screen
T4	☆☆☆	Review mention, text-mined	Review article

In report:

ATP6V1A drives lysosomal acidification [★★★: PMID:12345678].
It has been implicated in cancer metabolism [★☆☆: TCGA data].

根据证据强度对结论分级：

层级	符号	描述	示例
T1	★★★	有直接证据的机制研究	CRISPR敲除研究
T2	★★☆	功能研究	siRNA敲低
T3	★☆☆	关联/筛选结果	GWAS、高通量筛选
T4	☆☆☆	综述提及、文本挖掘	综述文章

报告中示例：

ATP6V1A驱动溶酶体酸化 [★★★: PMID:12345678]。
它与癌症代谢有关 [★☆☆: TCGA数据]。

Mandatory Completeness

强制完整性

All sections must exist, even if "data unavailable":

undefined

所有章节必须存在，即使显示“数据不可用”：

undefined

Pathogen Involvement

病原体关联

No pathogen interactions identified in literature or databases. Source: Literature search, UniProt annotations

undefined

在文献和数据库中未发现病原体互作。 来源：文献搜索、UniProt注释

undefined

Report Quality Metrics

报告质量指标

Quality	Description	Tool Calls	Sections
Excellent	Multi-database, evidence-graded	30+	All mandatory, detailed
Good	Cross-referenced, sourced	15-30	All mandatory, adequate
Adequate	Single-database focus	5-15	Core sections only
Poor	Single tool, no sources	<5	Incomplete

质量	描述	工具调用次数	章节
优秀	多数据库、证据分级	30+	所有必填章节，内容详细
良好	交叉引用、有来源	15-30	所有必填章节，内容充足
合格	单一数据库聚焦	5-15	仅核心章节
较差	单一工具、无来源	<5	内容不完整

Strategy 7: Use Specialized Skills for Specific Tasks

策略7：针对特定任务使用专业技能

CRITICAL: For specific research tasks, use specialized skills (not this general skill).

至关重要：对于特定研究任务，使用专业技能（而非本通用技能）。

Task-Specific Skill Selection

特定任务技能选择

Task	Recommended Skill
Data Retrieval
Chemical compounds	`tooluniverse-chemical-compound-retrieval`
Expression data	`tooluniverse-expression-data-retrieval`
Protein structure	`tooluniverse-protein-structure-retrieval`
Sequence retrieval	`tooluniverse-sequence-retrieval`
Research & Profiling
Disease research	`tooluniverse-disease-research`
Drug profiling	`tooluniverse-drug-research`
Literature review	`tooluniverse-literature-deep-research`
Target analysis	`tooluniverse-target-research`
Clinical Decision Support
Drug safety analysis	`tooluniverse-pharmacovigilance`
Precision oncology treatment	`tooluniverse-precision-oncology`
Rare disease diagnosis	`tooluniverse-rare-disease-diagnosis`
Variant interpretation	`tooluniverse-variant-interpretation`
Discovery & Design
Small molecule binder discovery	`tooluniverse-binder-discovery`
Drug repurposing	`drug-repurposing`
Protein therapeutic design	`tooluniverse-protein-therapeutic-design`
Outbreak Response
Infectious disease analysis	`tooluniverse-infectious-disease`
Infrastructure & Development
ToolUniverse installation/setup	`setup-tooluniverse`
Python SDK for AI scientist systems	`tooluniverse-sdk`

任务	推荐技能
数据检索
化学化合物	`tooluniverse-chemical-compound-retrieval`
表达数据	`tooluniverse-expression-data-retrieval`
蛋白质结构	`tooluniverse-protein-structure-retrieval`
序列检索	`tooluniverse-sequence-retrieval`
研究与分析
疾病研究	`tooluniverse-disease-research`
药物分析	`tooluniverse-drug-research`
文献综述	`tooluniverse-literature-deep-research`
靶点分析	`tooluniverse-target-research`
临床决策支持
药物安全性分析	`tooluniverse-pharmacovigilance`
精准肿瘤治疗	`tooluniverse-precision-oncology`
罕见病诊断	`tooluniverse-rare-disease-diagnosis`
变异体解读	`tooluniverse-variant-interpretation`
发现与设计
小分子结合物发现	`tooluniverse-binder-discovery`
药物重定位	`drug-repurposing`
蛋白质治疗药物设计	`tooluniverse-protein-therapeutic-design`
疫情响应
传染病分析	`tooluniverse-infectious-disease`
基础设施与开发
ToolUniverse安装/设置	`setup-tooluniverse`
AI科学家系统Python SDK	`tooluniverse-sdk`

When to Use This General Skill

何时使用本通用技能

Use this skill when:

Need general guidance on ToolUniverse usage
Task doesn't fit a specialized skill
Need to combine multiple specialized workflows
Exploring what's possible with ToolUniverse
Learning ToolUniverse best practices

在以下场景使用本技能：

需要ToolUniverse使用的通用指导
任务不适合专业技能
需要组合多个专业工作流
探索ToolUniverse的可用功能
学习ToolUniverse最佳实践

Strategy 8: Parallel Execution for Speed

策略8：并行执行以提升速度

CRITICAL: Run independent queries simultaneously for faster research.

至关重要：同时运行独立查询以加快研究速度。

When to Parallelize

何时并行执行

Parallel	Sequential
Different databases for same entity	Tool B needs output from Tool A
Multiple entities, same data type	Building an ID → using the ID
Independent research paths	Iterating through a list of results

并行	串行
同一实体的不同数据库查询	工具B需要工具A的输出
多个实体，相同数据类型	构建ID → 使用该ID
独立研究路径	遍历结果列表

Parallel Research Paths Example

并行研究路径示例

For target research, run these 8 paths simultaneously:

Identity - Names, IDs, sequence
Structure - 3D structure, domains
Function - GO terms, pathways
Interactions - PPI network
Expression - Tissue expression
Variants - Genetic variation
Drugs - Known drugs, druggability
Literature - Publications, trends

对于靶点研究，同时运行以下8条路径：

身份 - 名称、ID、序列
结构 - 3D结构、结构域
功能 - GO术语、通路
互作 - PPI网络
表达 - 组织表达
变异体 - 遗传变异
药物 - 已知药物、成药性
文献 - 出版物、趋势

Strategy 9: Iterative Completeness Check

策略9：迭代式完整性检查

CRITICAL: After gathering data, always ask "What else is missing?" to ensure comprehensive coverage.

至关重要：收集数据后，始终问自己“还缺少什么？”以确保全面覆盖。

The Completeness Loop

完整性循环

Gather initial data
    ↓
Review what you have
    ↓
Ask: "What aspects are still missing?"
    ↓
Identify gaps
    ↓
Search for tools to fill gaps
    ↓
Gather additional data
    ↓
Repeat until comprehensive

收集初始数据
    ↓
回顾已有内容
    ↓
提问：“哪些方面仍缺失？”
    ↓
识别缺口
    ↓
搜索填补缺口的工具
    ↓
收集额外数据
    ↓
重复直到全面

Universal Completeness Questions

通用完整性问题

After each research phase, ask:

Identity: Do I have all relevant identifiers and names?
Core data: Do I have the fundamental information for this entity type?
Context: Do I have surrounding/related information?
Relationships: Do I know what this connects to?
Variations: Do I know about variants, forms, or subtypes?
Evidence: Do I have supporting data from multiple sources?
Literature: Do I have recent publications on this topic?
Gaps: Have I documented what's unavailable?

在每个研究阶段后，提问：

身份：我是否拥有所有相关标识符和名称？
核心数据：我是否拥有该实体类型的基础信息？
上下文：我是否拥有相关的背景信息？
关系：我是否了解它的关联对象？
变异：我是否了解变异体、形式或亚型？
证据：我是否拥有来自多个来源的支持数据？
文献：我是否拥有该主题的最新出版物？
缺口：我是否记录了不可用的内容？

Gap-Filling Strategies

缺口填补策略

Gap Identified	Strategy
Missing data type	Search for tools with that data type
Single source only	Query additional databases
Outdated information	Check literature for recent updates
No experimental data	Look for predictions/computational data
Conflicting data	Find authoritative/curated sources
Shallow coverage	Dive deeper with specialized tools

识别的缺口	策略
缺少数据类型	搜索提供该数据类型的工具
仅单一来源	查询额外数据库
信息过时	检查文献获取最新更新
无实验数据	查找预测/计算数据
数据冲突	查找权威/curated来源
覆盖较浅	使用专业工具深入研究

When to Stop

何时停止

Stop the completeness loop when:

All relevant aspects have been addressed (even if "not found")
Multiple sources queried for key data
Gaps are documented, not ignored
No obvious missing pieces remain

当满足以下条件时停止完整性循环：

所有相关方面已覆盖（即使显示“未找到”）
关键数据已查询多个来源
缺口已记录，未被忽略
无明显缺失内容

Self-Review Questions

自我审查问题

Before finalizing any research:

Have I searched for ALL relevant tools?
Have I queried multiple databases?
Have I followed cross-references?
Have I checked recent literature?
Have I documented what's unavailable?
Is there any obvious gap I haven't addressed?
Would someone reading this ask "but what about X?"

在完成任何研究前：

我是否搜索了所有相关工具？
我是否查询了多个数据库？
我是否跟踪了交叉引用？
我是否检查了最新文献？
我是否记录了不可用的内容？
是否有任何明显缺口我未处理？
阅读报告的人是否会问“那X呢？”

Quick Reference: Tool Categories

快速参考：工具分类

Protein & Gene Tools

蛋白质与基因工具

UniProt, Proteins API, MyGene, Ensembl tools

UniProt, Proteins API, MyGene, Ensembl工具

Structure Tools

结构工具

RCSB PDB, PDBe, AlphaFold, InterPro tools

RCSB PDB, PDBe, AlphaFold, InterPro工具

Drug & Compound Tools

药物与化合物工具

ChEMBL, PubChem, DGIdb, ADMET-AI, DrugBank tools

ChEMBL, PubChem, DGIdb, ADMET-AI, DrugBank工具

Disease & Phenotype Tools

疾病与表型工具

OpenTargets, ClinVar, GWAS, HPO tools

OpenTargets, ClinVar, GWAS, HPO工具

Expression Tools

表达工具

GTEx, Human Protein Atlas, CELLxGENE tools

GTEx, Human Protein Atlas, CELLxGENE工具

Variant Tools

变异体工具

gnomAD, ClinVar, dbSNP tools

gnomAD, ClinVar, dbSNP工具

Pathway Tools

通路工具

Reactome, KEGG, WikiPathways, GO tools

Reactome, KEGG, WikiPathways, GO工具

Literature Tools

文献工具

PubMed, EuropePMC, OpenAlex, SemanticScholar tools

PubMed, EuropePMC, OpenAlex, SemanticScholar工具

Clinical Tools

临床工具

ClinicalTrials.gov, FAERS, PharmGKB, DailyMed tools

ClinicalTrials.gov, FAERS, PharmGKB, DailyMed工具

Troubleshooting Common Issues

常见问题排查

"Tool not found"

"工具未找到"

Search for similar tools using Tool_Finder
Check spelling of tool name
Try alternative tools for same data type

使用Tool_Finder搜索类似工具
检查工具名称拼写
尝试同一数据类型的替代工具

"Empty results"

"结果为空"

Check spelling of query terms
Try synonyms/aliases
Try alternative databases
Verify IDs are correct type

检查查询术语拼写
尝试同义词/别名
尝试替代数据库
验证ID类型是否正确

"Conflicting data"

"数据冲突"

Note all sources
Prefer curated databases
Document the conflict in report
Use evidence grading

记录所有来源
优先选择curated数据库
在报告中记录冲突
使用证据分级

"Incomplete picture"

"视图不完整"

Search for more tools
Query additional databases
Follow cross-references
Expand via literature

搜索更多工具
查询额外数据库
跟踪交叉引用
通过文献扩展

Strategy 10: English-First Tool Queries

策略10：工具查询优先使用英文

CRITICAL: Most ToolUniverse tools only accept English terms. Always translate queries to English before calling tools, regardless of the user's language.

至关重要：大多数ToolUniverse工具仅接受英文术语。无论用户使用何种语言，在调用工具前始终将查询转换为英文。

Language Handling Rules

语言处理规则

Default to English - All tool calls must use English search terms, entity names, and parameters
Translate non-English input - If the user's question is in Chinese, Japanese, Korean, or any other language, translate the relevant scientific terms to English before making tool calls
Respond in the user's language - While tools must be queried in English, deliver the final report/answer in the user's original language
Fallback to original language - Only if an English search returns no results, retry with the original-language terms
Check tool descriptions - A few tools may explicitly document multi-language support; use the original language only when the tool description says so

默认使用英文 - 所有工具调用必须使用英文搜索术语、实体名称和参数
翻译非英文输入 - 如果用户的问题是中文、日文、韩文或其他语言，在调用工具前将相关科学术语翻译为英文
用用户的语言回复 - 虽然工具调用必须使用英文，但最终报告/答案需使用用户的原始语言
** fallback到原始语言** - 仅当英文搜索无结果时，使用原始语言术语重试
查看工具描述 - 少数工具可能明确支持多语言；仅当工具描述说明时才使用原始语言

Examples

示例

User (Chinese): "研究EGFR靶点"
  → Tool calls: use "EGFR", "epidermal growth factor receptor" (English)
  → Report: deliver in Chinese

User (Japanese): "メトホルミンの安全性プロファイル"
  → Tool calls: use "metformin", "safety profile" (English)
  → Report: deliver in Japanese

User (Korean): "알츠하이머병 관련 유전자"
  → Tool calls: use "Alzheimer's disease", "associated genes" (English)
  → Report: deliver in Korean

用户（中文）："研究EGFR靶点"
  → 工具调用：使用"EGFR", "epidermal growth factor receptor"（英文）
  → 报告：用中文交付

用户（日文）："メトホルミンの安全性プロファイル"
  → 工具调用：使用"metformin", "safety profile"（英文）
  → 报告：用日文交付

用户（韩文）："알츠하이머병 관련 유전자"
  → 工具调用：使用"Alzheimer's disease", "associated genes"（英文）
  → 报告：用韩文交付

Why This Matters

为何这很重要

Scenario	Wrong Approach	Correct Approach
User asks in Chinese about "二甲双胍"	Pass "二甲双胍" to PubChem search	Translate to "metformin", search in English
User asks in Japanese about a disease	Pass Japanese disease name to OpenTargets	Translate to English disease name first
User asks in Spanish about a gene	Pass Spanish description to tool	Use standard gene symbol (e.g., TP53)

场景	错误做法	正确做法
用户用中文询问"二甲双胍"	将"二甲双胍"传入PubChem搜索	翻译为"metformin"，用英文搜索
用户用日文询问某疾病	将日文疾病名称传入OpenTargets	先翻译为英文疾病名称
用户用西班牙文询问某基因	将西班牙文描述传入工具	使用标准基因符号（如TP53）

Summary: The ToolUniverse Mindset

总结：ToolUniverse思维模式

Principle	Action
Clarify first	Confirm entity, scope, species, and output before researching
Search widely	10000+ tools; always discover more
Multi-hop persistence	5-10 tool calls per question is normal
Cross-reference	Query multiple databases for same data
Disambiguate first	Resolve IDs before research
Never give up	Fallbacks for every failure
Report comprehensively	Detail with sources and evidence grades
Use specialized skills	Apply domain-specific skills for focused tasks
Execute in parallel	Speed through concurrent execution
Check completeness	Ask "what's missing?" and fill gaps iteratively
English-first queries	Translate to English for tool calls; respond in user's language

The goal: Transform 10000+ tools into comprehensive, reliable scientific intelligence.

原则	行动
先澄清	研究前确认实体、范围、物种和输出形式
广泛搜索	10000+工具；始终发现更多工具
多跳持续查询	每个问题通常需要5-10次工具调用
交叉引用	针对同一数据查询多个数据库
先消除歧义	研究前解析ID
永不放弃	每个失败都有备选方案
全面报告	详细内容搭配来源和证据分级
使用专业技能	针对聚焦任务应用领域特定技能
并行执行	通过并发执行加快速度
检查完整性	问自己“还缺少什么？”并迭代填补缺口
优先英文查询	翻译为英文进行工具调用；用用户的语言回复

目标：将10000+工具转化为全面、可靠的科学智能。