PubMed quickstart
Type a plain-English query, read the matching titles and abstracts off the canvas, then ask the model to pull the genes, compounds and diseases it names.
You want to find the published literature on a target, mutation, or drug and mine it for entities; this searches PubMed, then asks the model to extract entities from the returned abstracts in a few steps.
PubMed Quickstart
Search
In chat, type a natural-language query:
The platform translates this into an E-utilities query and returns matching publications with titles, abstracts, and PubMed links. PubMed does not provide citation counts; for those, ask Ivy to look the papers up in Semantic Scholar.
Review Results
Results appear on the canvas showing:
- Publication titles and authors
- Journal and publication date
- Abstract previews
- PubMed links
Extract entities
Ask the model to mine the results — it reads the returned abstracts directly, no separate tool call:
The model extracts:
- Proteins and genes mentioned (e.g., EGFR, T790M, C797S)
- Compounds referenced (e.g., osimertinib, rociletinib)
- Diseases discussed
- Relationships between entities
Build a Literature Passport
For a comprehensive analysis:
This orchestrates PubMed search + entity extraction + trend analysis into a single report.
Next Steps
- Investigate a target: Send extracted protein names to Target Passport
- Check compound safety: Send extracted drug names to Safety Passport
- Find structures: Chain extracted proteins into ESMFold or PDB Search
Read next
Was this page helpful?
PubMed Search outputs
Read pmid, title, abstract, authors and journal per article — plus the doi, pmcid and mesh_terms that only pubmed_search returns, never advanced search.
PubMed Search troubleshooting
Zero results, a 45-second timeout, NCBI rate limits, and the missing abstract that advanced search never fetches — plus why you filter by date, not sort.