BARITONE - Adult Rare Autoimmune Diseases (ARADs) Health Knowledge Graph

This repository provides a health knowledge graph derived from the co-occurrence of normalized clinical concepts automatically extracted by Named Entity Recognition (NER) models from clinical case documents in the rheumatology domain. The clinical entities were manually normalized by assigning identifiers from a controlled medical terminology — in this case, SNOMED CT. The extracted concepts belong to two main semantic categories: diseases and symptoms.

The repository includes a .csv file containing the list of co-occurring entity pairs and their associated metadata. The file is organized into the following columns:

span_mention_1 – Text span of the first identified mention (original text).

normalized_entity_1 – Controlled vocabulary concept mapped to normalize mention 1.

code_mention_1 – Terminology identifier corresponding to mention 1 (e.g., SNOMED CT code).

mention_controlled_vocab – Terminology used for normalization of mention 1.

mention1_category – Semantic category (e.g., NER label) of mention 1.

mention1_norm_freq – Absolute frequency of the normalized mention 1.

span_mention_2 – Text span of the second identified mention (original text).

normalized_entity_2 – Controlled vocabulary concept mapped to normalize mention 2.

code_mention_2 – Terminology identifier corresponding to mention 2 (e.g., SNOMED CT code).

mention_controlled_vocab – Terminology used for normalization of mention 2.

mention2_category – Semantic category (e.g., NER label) of mention 2.

mention2_norm_freq – Absolute frequency of the normalized mention 2.

co-occurrence – Number of co-occurrences computed between mention 1 and mention 2 across the dataset.

This dataset supports the visualization of a rheumatology-specific biomedical knowledge graph facilitating the discovery of novel relationships between rheumatologic diseases and other medical conditions that are not frequently associated, as well as the identification of atypical symptom manifestations differing from classical clinical presentations.

Contact

If you have any questions or suggestions, please contact us at:

- Miguel Rodríguez Ortega (<miguel [dot] rod [at] bsc [dot] com>)