/literature-review

Automated literature review with clinical trials data.

Primary LanguageJupyter Notebook

Literature Review

Todo

  • When generating schema, we want to query around 1000-5000 trials. This might differ from the num_trials specified by the user.
  • Don't insert study if it already exists.
  • Modify _create_table algorithm. Right now, it doesn't work with PostgreSQL because we cannot define foreign relations before tables are created.

Notebooks

  1. NLP Notes
  2. Summary Statistics
  3. Embeddings
  4. Vector Database

Installation

Install Poetry

If poetry is not installed,

pip install pipx
pipx install poetry

Install Repo with Poetry

git clone https://github.com/clinical-trials-research/literature-review.git
cd literature-review
poetry install

API

  • ClinicalTrials(*, num_studies=1000, connection=None, schema_directory="schema.json") - Creates ClinicalTrials object which provides an interface for database management.

    • num_studies - Number of studies to retrieve in each call to update_database or get_studies.

    • connection - Database connection. If not specified, creates new one.

    • schema_directory - Schema to use. If not specified, creates a schema.

      from litreview import ClinicalTrials

trials = ClinicalTrials()

  • ClinicalTrials.query(*fields) - Query the database for the specified fields

    trials.query("NCTId", "BriefDescription")

  • ClinicalTrials.update_database() - Update the database with num_studies number of studies.

    trials.update_database()

  • ClinicalTrials.get_studies() - Retrieve num_studies number of studies in a list.

    studies = trials.get_studies()

Resources

Clinical Trials

NLP