跳到正文

医学公开数据库查询与登记

使用有明确接口的公开数据库查询方法保存请求、筛选、分页与来源,可用于医学文献、临床试验及药物资料检索。

开源精选医疗业务86 个文件50,135 Stars浏览次数 0K-Dense-AI/scientific-agent-skills MIT版本 a1043ef
开始使用

下载前阅读 SKILL.md 中的使用条件,并确认所用工具支持对应能力。

输入资料
指定数据库和查询问题 实际接口、筛选条件及使用范围 阅读本节完整说明
交付内容
可复查的查询资料 请求、分页与来源记录 阅读本节完整说明
  1. 获取完整包,保留配套文件的目录结构。
  2. 按照原文准备所需工具、依赖与输入资料。
  3. 阅读对应使用步骤与输入输出说明,在自己的环境中验证结果。

使用提示摘自原文,配套文件保留来源版本;实际运行条件以各包说明为准。

正文

获取原文件

references/geo.md

Base URLs

Purpose URL
E-utilities https://eutils.ncbi.nlm.nih.gov/entrez/eutils/
GEO direct query https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi

Important: The database name is gds

The Entrez database for GEO is gds (not geo). It contains all GEO record types: GDS datasets, GSE series, GPL platforms, GSM samples. Filter by type using [ETYP] in the search term.

Key Endpoints

eSearch — Search GEO

GET /esearch.fcgi?db=gds&term={query}&retmode=json&retmax={n}

Parameters:

  • db=gds (required)
  • term — search query with field tags
  • retmax — max results (default 20)
  • retstart — offset for pagination
  • retmode=json — get JSON response
  • usehistory=y — store results server-side for large queries
  • api_key — NCBI API key (optional, raises rate limit)

Entry type filters ([ETYP])

  • gds[ETYP] — curated GEO DataSets
  • gse[ETYP] — GEO Series (most common, use this by default)
  • gpl[ETYP] — platforms
  • gsm[ETYP] — samples

Other field tags

  • [Organism] — e.g. "Homo sapiens"[Organism]
  • [PDAT] — publication date
  • [Title] — title search
  • Boolean: AND, OR, NOT (uppercase)

Example — cancer GSE series in human:

/esearch.fcgi?db=gds&term=cancer+AND+gse[ETYP]+AND+"Homo+sapiens"[Organism]&retmax=10&retmode=json

Response:

{
  "esearchresult": {
    "count": "15432",
    "retmax": "10",
    "idlist": ["200012345", "200067890"],
    "querytranslation": "cancer AND gse[ETYP]"
  }
}

The IDs returned are numeric Entrez UIDs, not accessions. Resolve them with ESummary and read accession; do not depend on an arithmetic UID conversion.

eSummary — Get metadata for UIDs

GET /esummary.fcgi?db=gds&id={uid_list}&retmode=json

Key response fields per record:

  • accession — e.g. "GSE12345"
  • title, summary
  • taxon — organism
  • entrytype — "GDS", "GSE", "GPL", "GSM"
  • gdstype — e.g. "Expression profiling by array"
  • n_samples — sample count
  • pubmedids — linked PubMed IDs
  • pdat — publication date
  • samples — array of sample objects
  • ftplink — data download path

Example:

/esummary.fcgi?db=gds&id=200012345&retmode=json

GEO Direct Query — Full records by accession

GET https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc={accession}&form={format}&view={detail}

Parameters:

  • acc — GEO accession (GSE12345, GDS1234, GPL570, GSM12345)
  • targ — self, gsm (samples), gpl (platform), gse (series)
  • form — text (SOFT format), xml (MINiML), html
  • view — quick, brief, full, data

Example — series metadata in SOFT:

acc.cgi?acc=GSE53757&targ=self&form=text&view=brief

Note: acc.cgi does not return JSON. Use eSearch + eSummary for JSON results. Use acc.cgi when you need full SOFT/MINiML records.

GET /elink.fcgi?dbfrom=gds&db=pubmed&id={uid}&retmode=json

Practical Workflow

For most queries, use this two-step approach:

  1. eSearch to find UIDs matching the query
  2. eSummary to get metadata for those UIDs

This gives you JSON throughout.

Important Notes

  • GDS records are mostly frozen — NCBI stopped curating new GDS. Use gse[ETYP] for comprehensive results.
  • eFetch has limited support for the gds database. Use eSummary for metadata or acc.cgi for full records.
  • URL-encode spaces as + and quotes as %22.

Rate Limits

  • Without API key: 3 requests/second
  • With API key: 10 requests/second (free registration at ncbi.nlm.nih.gov/account/settings)
  • Include &email=user@example.com as a courtesy
  • For large result sets, use the History server (usehistory=y then pass WebEnv and query_key to eSummary)

CONTACT HAONAN

联系haonan

医疗 AI · 产品与独立开发

刘浩南

关注医疗 AI 从 demo 到真实落地,聚焦产品化、商业化与院内工作流效率提升。

12年医疗产研经验,大湾区头部医疗AI公司前产品总监,独立开发者

医疗 AI 产品化医疗 AI 商业化院内工作流效率
刘浩南的微信联系二维码
微信扫一扫,联系 haonan