OntoMedRec

The model repository of OntoMedRec. To run the model, please make sure you have the following packages:

• PyTorch >= 1.12.0
• LTNTorch
• Numpy
• Pandas
• sklearn

Data Preprocessing

To get access to MIMIC dataset, please follow the instruction on https://physionet.org/content/mimiciii/1.4/

Download the raw DDI dataset as per the instruction in https://github.com/ycq091044/SafeDrug

Please put the diagnoses, procedures, prescriptions and admissions datasets to the ./data directory.

To get the training, testing and validation records, please and run the following commands:

cd data
python processing.py

Running the fine-tuning for OntoMedRec and other baselines

We prepared the pretrained embeddings of OntoMedRec and other baselines. To fine-tune the model, you can simply run the following command

python src/downstream/MICRON.py --embd_mode omr --pro_taxo --epochs 60

Other options of the --embd_mode parameters are gat, gcn and random. --pro_taxo parameter means using the OntoMedRec representation for procedures.

To test the performance of the fine-tuned model, simply add --test to the above command.

python src/downstream/MICRON.py --embd_mode omr --pro_taxo --test --epochs 60

Acknowledgement

We would like to express our sincere gratitute to the repective authors of the following papers their code base:

• https://github.com/BarryRun/COGNet
• https://github.com/ycq091044/SafeDrug
• https://github.com/Melinda315/4SDrug