SuperAPI AI Directory

pubmed-systematic-search

General & other

View repository

Updated 10-16 09:30

About this skill

Systematic literature search

Description

This workflow is used in situations where you want to "search comprehensively" rather than "take a quick look first." The core approach is to first use generate_search_queries to obtain MeSH terms, synonyms, and suggested queries, then have the Agent or user assemble them into an explicit Boolean query, and finally execute it with unified_search.

Trigger conditions

  • "systematic search"
  • "comprehensive search"
  • "literature review"
  • 「comprehensive search」
  • 「systematic review」
  • Mentions MeSH, synonym expansion, search strategy

Correct workflow

generate_search_queries
→ 整理 MeSH / 同義詞 / suggested_queries
→ 手動或由 Agent 組 Boolean 查詢
→ analyze_search_query
→ unified_search
→ fetch_article_details / prepare_export / save_pipeline

There is currently no public standalone merge tool workflow. Every unified_search itself performs multi-source integration and deduplication; if you run multiple rounds of strategies, the approach should be to compare the results of each round, adjust queries, or save the process as a pipeline, rather than relying on the old merge approach.


Step 1: Obtain search materials

generate_search_queries(
    topic="remimazolam ICU sedation",
    strategy="comprehensive"
)

strategy options

  • comprehensive: default, suitable for comprehensive search
  • focused: converge to a higher evidence level
  • exploratory: relax, find more variants and synonyms

The fields you actually need to use

  • mesh_terms: standard terms and corresponding synonyms
  • all_synonyms: can be directly assembled into OR groups
  • suggested_queries: use as reference, not the final answer
  • pubmed_translation: check how PubMed actually interprets the query

Step 2: Assemble the Boolean query

Example: assemble an executable query from the materials

query = '''
("Intensive Care Units"[Title/Abstract] OR ICU[Title/Abstract] OR "critical care"[Title/Abstract])
AND
(remimazolam[Title/Abstract] OR "CNS 7056"[Title/Abstract] OR "ONO 2745"[Title/Abstract])
AND
(sedation[Title/Abstract] OR "procedural sedation"[Title/Abstract])
'''

Two principles

  1. Main concepts are usually joined with AND
  2. Synonyms and aliases are usually joined with OR

Step 3: Analyze before executing

analyze_search_query(query=query)

This step is used to confirm:

  • Whether the query is too broad or too narrow
  • Whether the PubMed translation matches expectations
  • Whether there are spelling errors or misplaced concepts

Step 4: Execute the search

unified_search(
    query=query,
    sources="pubmed,europe_pmc,openalex",
    limit=50,
    ranking="quality",
    filters="year:2020-2025, species:humans, clinical:therapy",
    output_format="json"
)

Common adjustment methods

  • Want more comprehensive: add options="preprints"
  • Want faster: add options="shallow"
  • Do not automatically relax: add options="no_relax"
  • Value recency:ranking="recency"
  • Value evidence quality:ranking="quality"

Complete example

Scenario: comprehensive search for remimazolam ICU sedation

# Step 1: 取得 MeSH 與同義詞素材
materials = generate_search_queries(
    topic="remimazolam ICU sedation",
    strategy="comprehensive"
)

# Step 2: 組裝查詢
query = '''
("intensive care"[Title/Abstract] OR ICU[Title/Abstract] OR "critical care"[Title/Abstract])
AND
(remimazolam[Title/Abstract] OR "CNS 7056"[Title/Abstract] OR "ONO 2745"[Title/Abstract])
AND
(sedation[Title/Abstract] OR "procedural sedation"[Title/Abstract])
'''

# Step 3: 先分析
analyze_search_query(query=query)

# Step 4: 再執行
unified_search(
    query=query,
    limit=50,
    ranking="quality",
    filters="year:2020-2025, species:humans, clinical:therapy",
    output_format="json"
)

If there are too few or too many results

Too few results

  • Change strategy to exploratory
  • Remove some restriction conditions
  • Reduce AND, add synonym OR
  • Remove clinical or overly narrow age/gender restrictions

Too many results

  • Change strategy to focused
  • Add topic qualifiers
  • Add filters="year:..., clinical:..."
  • Converge sources to pubmed,europe_pmc

When reusable

When the search strategy has stabilized, do not reassemble it every time:

save_pipeline(name="icu_sedation_review", pipeline_config="...")

Afterwards you can:

unified_search(pipeline="saved:icu_sedation_review")