PPubMed10 Apr 2024Cited 6×

NetMe 2.0: a web-based platform for extracting and modeling knowledge from biomedical literature as a labeled graph

Antonio Di Maria, Lorenzo Bellomo, Fabrizio Billeci, Alfio Cardillo, Salvatore Alaimo, Paolo Ferragina, Alfredo Ferro, Alfredo Pulvirenti, Macha Nikolski

Abstract

To assess the quality of the BKG built by NetMe 2.0, we conducted an experimental evaluation based on manually curated GDA gathered from DisGeNET. Therefore, we designed four case studies to assess NetMe 2.0 accuracy. The first evaluation involves queries guided by paper IDs specified by the user, thus evaluating the effectiveness of knowledge extraction from a set of papers via the comparison with the GDAs present in them and annotated in DisGeNET. The second case study queries NetMe 2.0 with gene names, like in PubMed, to evaluate its ability to infer the same GDAs found by experts in DisGeN

A figure from NetMe 2.0: a web-based platform for extracting and modeling knowledge from biomedical literature as a labeled graph — fig. from the paper

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

NetMe 2.0 is a web platform that builds on‑the‑fly biomedical knowledge graphs from texts and uses Graph‑RAG to generate summarizing sentences, achieving higher precision‑recall than prior tools.

What’s new

i.OntoTagMe, a customized Wikidata‑based entity linker covering ~3 million biomedical entities
ii.A SpaCy‑based relationship‑inference module for extracting edges between entities
iii.An on‑the‑fly Graph‑RAG module that summarizes graph paths using OpenAI GPT‑3.5
iv.A visual web interface with network‑analysis tools and Docker deployment

How they did it

The platform first extracts biomedical entities from user‑provided texts using OntoTagMe, then infers relations between those entities with a SpaCy dependency parser to build a labeled knowledge graph. Users can explore the graph via an AngularJS/CytoscapeJS GUI and invoke the Graph‑RAG module, which retrieves sentences along selected paths and sends them to OpenAI GPT‑3.5 to generate concise summaries.

What they found

Only 2 of 46 edges were missed in the first case study, with 8 misses due to missing evidence
Recall values between 0.58 and 0.77 when extracting known gene‑disease associations
NetMe 2.0 achieved 87 correct edges versus 63 for NetMe 1.0 (≈38% improvement, 24 additional correct relations)
Outperformed other BKG builders with absolute improvements ranging from 2% to 87%

Caveats

✕Current annotation quality may be improved by integrating the UMLS Metathesaurus (future work)
✕The system has been evaluated on a limited set of case studies and datasets
✕Graph‑RAG can retrieve ambiguous or uncertain sentences, requiring manual verification

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

Similar papers in all four archives

The same idea searched in arXiv, PubMed, bioRxiv and chemRxiv at once, so a method can lead you to where it is being applied.

AarXiv4 found

PPubMed3 found

BbioRxiv4 found

CchemRxiv4 found

Searching the other archives…