BRONCO

BRONCO provides a curated corpus for extracting gene-variant-disease-drug relations from biomedical literature focused on cancer and anti-tumor drug screening to support variant curation and BioNLP method development.

Key Features:

• Corpus contents: Over 400 annotated variants with explicit relations to genes, diseases, drugs, and cell lines.
• Source material: Manual extraction and annotation from 108 full-text articles.
• Annotation scope: Gene-variant-disease-drug relation annotations tailored to cancer and anti-tumor drug screening contexts.
• Gold-standard dataset: Curated as a benchmark resource for training and evaluating biomedical natural language processing (BioNLP) methods.
• Method evaluation: Employed for quantitative and qualitative evaluation of three state-of-the-art BioNLP methods.
• Post-processing development: Supported creation of post-processing modules that significantly improved BioNLP method performance.
• Motivation: Addresses the increased volume of genomic variant literature driven by next-generation sequencing technologies.

Scientific Applications:

• BioNLP method training and benchmarking: Provide a gold-standard reference for training and assessing named-entity recognition and relation extraction systems for variants, genes, diseases, and drugs.
• Variant information extraction: Facilitate automated extraction of functional variant information from cancer literature for downstream analysis.
• Precision medicine research: Support curation and validation of variant–drug and variant–disease associations relevant to precision oncology.
• Cancer genomics and drug screening studies: Enable evaluation of text-mining approaches applied to anti-tumor drug screening and related cancer research topics.

Methodology:

Relations were manually extracted and annotated from 108 full-text articles to produce a corpus of over 400 annotated variants, which was used to quantitatively and qualitatively evaluate three BioNLP methods and to develop post-processing modules.