metadata

license: mit
datasets:
  - bigbio/biored
language:
  - en
metrics:
  - f1

Model Card for BioNExt

BioNExt, is an end-to-end Biomedical Relation Extraction and Classifcation system. The work utilized three modules, a Tagger (Named Entity Recognition), Linker (Entity Linking) and an Extractor (Relation Extraction and Classification).

This repositories contains two models:

Tagger: Named Entity Recognition module, which performs 6 class biomedical NER: Genes, Diseases, Chemicals, Variants (mutations), Species, and Cell Lines.
Extractor: Performs Relation Extraction and classification. The classes for the relation Extraction are: Positive Correlation, Negative Correlation, Association, Binding, Drug Interaction, Cotreatment, Comparison, and Conversion.

For a full description on how to utilize our end-to-end pipeline we point you towards our GitHub repository.

Model Details

Model Description

Developed by: IEETA
Model type: BERT Base
Language(s) (NLP): English
License: MIT
Finetuned from model: BioLinkBERT-Large

Model Sources

Repository: IEETA BioNExt GitHub
Paper: Towards Discovery: An End-to-End System for Uncovering Novel Biomedical Relations [Awaiting Publication]

Authors:

Tiago Almeida (ORCID: 0000-0002-4258-3350)
Richard A A Jonker (ORCID: 0000-0002-3806-6940)
Rui Antunes (ORCID: 0000-0003-3533-8872)
João R Almeida (ORCID: 0000-0003-0729-2264)
Sérgio Matos (ORCID: 0000-0003-1941-3983)

Uses

Note we do not take any liability for the use of the model in any professional/medical domain. The model is intended for academic purposes only.

How to Get Started with the Model

Please refer to our GitHub repository for more information on our end-to-end inference pipeline: IEETA BioNExt GitHub

Training Details

Training Data

The training data utilized was the BioRED corpus, wihtin the scope of the BioCreative-VIII challenge.

Ling Luo, Po-Ting Lai, Chih-Hsuan Wei, Cecilia N Arighi, Zhiyong Lu, BioRED: a rich biomedical relation extraction dataset, Briefings in Bioinformatics, Volume 23, Issue 5, September 2022, bbac282, https://doi.org/10.1093/bib/bbac282

Results

As evaluated as an end to end system, our results are as follows:

Tagger: 43.10
Linker: 32.46
Extractor: 24.59

Configuration	Entity Pair (P/R/F%)	+ Relation (P/R/F%)	+ Novel (P/R/F%)
Competition best	-/-/55.84	-/-/43.03	-/-/32.75
BioNExt (end-to-end)	45.89/40.63/43.10	34.56/30.60/32.46	26.18/23.18/24.59

Citation

BibTeX:

[Awaiting Publication]