meng-ma-biomed-AI/AutoRD

AutoRD

• AutoRD: An Automatic and End-to-end Rare Disease Knowledge Graph Construction Framework Based on Ontologies-enhanced Large Language Models

Quick Start

• 0. put your OPENAI_API_KEY at env.py
• 0. get the latest ontolgies on the Internet. HOOM_en_2.0.owl, ORDO_en_4.3.owl, mondo.obo
• 0. get the test dataset RareDis-v1 from the authors of the original paper

# data preprocessing
cd data_preprocessing

# put original dataset RareDis at `data_preprocessing/data` and rename it and cover it at `data_preprocessing/data/RareDis-fixed`

# fix some data annotation errors
cd RareDis-fixed
python fix_data.py

cd ../
python generate_data.py
python parse_HOOM.py
python parse_mondo_obo.py
python parse_ORDO.py


# put new generated dataset and processed ontologies at `./data`

# run main code
cd ../
bash run.sh

• Find your results at ./cache_data

Baselins

## run baselines
# fine-tuning baseline
cd finetuning_baseline
python main.py

# pure LLM baseline
cd LLM_baseline
python main.py