README.md

metadata

dataset_info:
  config_name: OptimALBaselineDataset
  features:
    - name: _unit_id
      dtype: int64
    - name: Worker Answer
      dtype: string
    - name: context
      dtype: string
    - name: drug_name
      dtype: string
    - name: disease_name
      dtype: string
  splits:
    - name: train
      num_bytes: 155844
      num_examples: 187
  download_size: 53627
  dataset_size: 155844
configs:
  - config_name: OptimALBaselineDataset
    data_files:
      - split: train
        path: OptimALBaselineDataset/train-*
task_categories:
  - text-classification
language:
  - en
tags:
  - medical
  - biology
  - drugs
  - relation-extraction
  - relation-classification
  - disease
pretty_name: OptimAL
size_categories:
  - n<1K

Dataset Card for OptimAL

Dataset Description

• Repository: https://github.com/MaastrichtU-IDS/OptimAL
• Paper: Relation extraction from DailyMed structured product labels by optimally combining crowd, experts and machines

Dataset Summary

OptimAL Dataset Summary:

The dataset in this repository is relevant to drug discovery and clinical decision support. The authors of the dataset reported on the combination of weak supervision (i.e., programmatic labeling and crowdsourcing) and deep learning methods for relation extraction from DailyMed text to create a higher quality drug-disease relation dataset. The generated drug-disease relation data showed a high overlap with DrugCentral, a manually curated dataset. Using this dataset, they constructed a machine learning model to classify relations between drugs and diseases from text into four categories; treatment, symptomatic relief, contradiction, and effect.

The data that is available in this repository is the OptimALBaselineDataset: https://github.com/MaastrichtU-IDS/OptimAL/blob/master/data/OptimALBaselineDataset.csv

Languages

The language in the dataset is English.

Dataset Structure

Dataset Instances

An example of 'train' looks as follows:

{
  "_unit_id": 2270472226,
  "Worker Answer": "effect",
  "context": "(See INDICATIONS AND USAGE and WARNINGS.)  Experience in over 1,400 patients with nifedipine immediate-release capsules in a noncomparative clinical trial has shown that concomitant administration of nifedipine and beta-blocking agents is usually well tolerated, but there have been occasional literature reports suggesting that the combination may increase the likelihood of congestive heart failure, severe hypotension, or exacerbation of angina.",
  "drug_name": "Nifedipine",
  "disease_name": "CONGESTIVE HEART FAILURE"
}

Data Fields

• _unit_id: the unique identifier for this data entry, an int64 feature.
• Worker Answer: the answer or classification provided by the worker based on the context, a string feature.
• context: the text providing the scenario or situation where the drug and disease interaction is mentioned, a string feature.
• drug_name: the name of the drug discussed in the context, a string feature.
• disease_name: the name of the disease discussed in relation to the drug within the context, a string feature.

Citation

BibTeX: @article{SHINGJERGJI2021103902, title = {Relation extraction from DailyMed structured product labels by optimally combining crowd, experts and machines}, journal = {Journal of Biomedical Informatics}, volume = {122}, pages = {103902}, year = {2021}, issn = {1532-0464}, doi = {https://doi.org/10.1016/j.jbi.2021.103902}, url = {https://www.sciencedirect.com/science/article/pii/S1532046421002318}, author = {Krist Shingjergji and Remzi Celebi and Jan Scholtes and Michel Dumontier}, keywords = {Drug-disease relation classification, Drug indications, Drug data quality, Drug repositioning, Weak supervision, Programmatic labeling, Crowdsourcing, Human-in-the-loop, Machine learning}, abstract = {The effectiveness of machine learning models to provide accurate and consistent results in drug discovery and clinical decision support is strongly dependent on the quality of the data used. However, substantive amounts of open data that drive drug discovery suffer from a number of issues including inconsistent representation, inaccurate reporting, and incomplete context. For example, databases of FDA-approved drug indications used in computational drug repositioning studies do not distinguish between treatments that simply offer symptomatic relief from those that target the underlying pathology. Moreover, drug indication sources often lack proper provenance and have little overlap. Consequently, new predictions can be of poor quality as they offer little in the way of new insights. Hence, work remains to be done to establish higher quality databases of drug indications that are suitable for use in drug discovery and repositioning studies. Here, we report on the combination of weak supervision (i.e., programmatic labeling and crowdsourcing) and deep learning methods for relation extraction from DailyMed text to create a higher quality drug-disease relation dataset. The generated drug-disease relation data shows a high overlap with DrugCentral, a manually curated dataset. Using this dataset, we constructed a machine learning model to classify relations between drugs and diseases from text into four categories; treatment, symptomatic relief, contradiction, and effect, exhibiting an improvement of 15.5% with Bi-LSTM (F1 score of 71.8%) over the best performing discrete method. Access to high quality data is crucial to building accurate and reliable drug repurposing prediction models. Our work suggests how the combination of crowds, experts, and machine learning methods can go hand-in-hand to improve datasets and predictive models.} } APA: - Shingjergji, K., Celebi, R., Scholtes, J., & Dumontier, M. (2021). Relation extraction from DailyMed structured product labels by optimally combining crowd, experts and machines. Journal of Biomedical Informatics, 122, 103902. https://doi.org/10.1016/j.jbi.2021.103902

Dataset Card Authors

@phucdev