Knowledge graph informed AI for disease research
A framework for applying best in class knowledge graph and retrieval augmented generation techniques to biomedical data and user queries. This project harnesses code from papers in a way that makes comparison on test data across different approaches possible. It offers portable Python tooling to make building and observing your AI experiments a breeze.
This Python package can be installed locally. First, clone this repository:
$ git clone git@github.com:keppy/disease-lab.git
For now you will need poetry installed to build the package: https://python-poetry.org
Once Poetry is installed, you can install this package locally by running:
$ poetry install
Next, in your Python notebook or script, import the entity extractor:
from disease_lab.entity_extraction.disease_query import expand_disease_query
Currently you can run the interactive command line tool and extract diseases from a text input string
To install dependencies and start the CLI:
$ poetry install
$ disease-lab
If you want to use an open source model, make sure Ollama is installed, updated, and running, and then execute:
$ disease-lab llama3
You will be presented with a prompt > , enter a string containing multiple diseases to get back the entities. Note that this will not return other entities like genes or drug compounds.
https://arxiv.org/abs/2311.17330
https://arxiv.org/abs/2405.14831
TODO: give credit to specific techniques from papers and isolate them for testing