Overview
Linked Discoveries, from the US National Library of Medicine (NLM), uses a version of the biomedical language model BiomedBERT to build a “neighbourhood” of articles closely related to a PubMed article, which can include replication studies. The model compares the titles, abstracts and keywords of PubMed records to judge how closely they are related.
A neighbourhood shows the 50 most similar records by default, and users can expand it to up to 200. Users can explore the neighbourhood in graph, timeline and table views, and narrow it by the conditions, genes and chemicals the articles share. It also shows context such as citation relationships, reviews, retractions and NIH funding, drawn from PubMed and other NLM databases.
NLM launched Linked Discoveries on 24 September 2026 as an experimental pilot covering more than 29 million PubMed publications. Users reach it through a button on PubMed abstract pages. It is an early product of the National Institutes of Health’s initiative to strengthen replication and reproducibility in the research it funds.
Notes and quotes
- NIH says the tool does not judge the quality of a study or determine whether a finding has been replicated.
- “Rather than treating published literature as a static collection of facts, Linked Discoveries enables dynamic and discerning engagement.” Peter Embí, Director, National Library of Medicine.
Sources
- NIH launches new PubMed tool to strengthen research replication and reproducibility, National Institutes of Health, 24 September 2026
- The National Library of Medicine Introduces Linked Discoveries: A New Way to Explore Connections in Biomedical Literature, National Library of Medicine, 24 September 2026
- Introducing Linked Discoveries, NLM Technical Bulletin, 24 September 2026
- Linked Discoveries User Guide, National Library of Medicine
This entry was written with the help of AI, from the sources listed. Please check the sources before relying on it, and let us know if anything needs correcting.
