Implementation of paper: Hyper-relational Knowledge Enhanced Network for Hypertension Medication Recommendation

Folder Specification

• DATA_prepare/
  • patient_hyp.ipynb: Extract the EHR of patients with hypertension.
  • med_hyp.ipynb: Extract the antihypertensive medications used by patients with hypertension.
  • ndc2RXCUI.txt: NDC to xnorm mapping file
  • RXCUI2atc4.csv: NDC code to ATC4 level code mapping file
• DR_module/
  • data/ (For a fair comparision, we use the same data and pre-processing scripts used in Safedrug)
    • input/ mapping files that collected from external sources
      • PRESCRIPTIONS.csv: the prescription file from MIMIC-IV_2.2 dataset
      • DIAGNOSES_ICD.csv: the diagnosis file from MIMIC-IV_2.2 dataset
      • PROCEDURES_ICD.csv: the procedure file from MIMIC-IV_2.2 dataset
      • RXCUI2atc4.csv: xnorm code to ATC4 code mapping file. This file is obtained from https://github.com/sjy1203/GAMENet.
      • ndc2RXCUI.txt: NDC to xnorm mapping file. This file is obtained from https://github.com/sjy1203/GAMENet.
      • drug-atc.csv: drug to atc code mapping file. This file is obtained from https://github.com/sjy1203/GAMENet.
      • drug-DDI.csv: this a large file, could be downloaded from https://drive.google.com/file/d/1mnPc0O0ztz0fkv3HF-dpmBb8PLWsEoDz/view?usp=sharing
      • drugbank_drugs_info.csv: drug information table downloaded from drugbank here https://www.dropbox.com/s/angoirabxurjljh/drugbank_drugs_info.csv?dl=0, which is used to map drug name to drug SMILES string.
      • ICD_tensor.csv: This is the medical entity vector file obtained by training HKG through KG_module. Users can use it directly.
    • output/ other files that generated from mapping files and MIMIC dataset (we attach these files here, user could use our provided scripts to generate)
      • atc3toSMILES.pkl: drug ID (we use ATC-3 level code to represent drug ID) to drug SMILES string dict
      • records_final.pkl: The final diagnosis-procedure-medication EHR records of each patient, used for train/val/test split. Under MIMIC Dataset policy, we are not allowed to distribute the datasets. Practioners could go to https://physionet.org/content/mimiciv/2.2/ and requrest the access to MIMIC-III dataset and then run our processing script to get the complete preprocessed dataset file.
      • ddi_A_final.pkl: ddi adjacency matrix
      • ehr_adj_final: EHR adjacency matrix
      • ddi_mask_H.pkl: H mask structure
      • voc_final.pkl: diag/prod/med index to code vocabulary
      • substructure_smiles.pkl: med substructure to SMILES mapping
    • dataset processing scripts
      • processing.py
      • get_SMILES.py
  • src/
    • HKGRec.py: train/test HKGRec
    • models.py: full model of HKGRec and baseline models
    • util.py: setting file
    • layer.py: setting file
• KG_module/
  • data/
    • n-ary-1.11.json: the hyper-relational facts.
    • data.ipynb: convert hyper-relational facts into learnable n-ary form.
    • processing.pkl: divided into test set, training set, validation set.
  • other files
    • batching.py: replace the tail entity to get negative facts.
    • builddata.py: build facts data.
    • model.py: full model of knowledge_driven encoder.
    • KG.py: knowledge_driven encoder.
    • entity_id.txt: entities in the constructed HKG to ID mapping file.
    • relation_id.txt: relations in the constructed HKG to ID mapping file.

Step 1: Data Extract

• Go to https://physionet.org/content/mimiciv/2.2/ to download the MIMIC-IV_2.2 dataset (You may need to get the certificate)
• go into the folder and unzip three main files (PROCEDURES_ICD.csv.gz, PRESCRIPTIONS.csv.gz, DIAGNOSES_ICD.csv.gz)
• pip install jupyter notebook
  • run patient_hyp.ipynb, get the EHR of patients with hypertension step by step.
  • run med_hyp.ipynb, get the antihypertensive medications used by hypertension patients step by step.

Step 2: Data Processing

• change the path in processing.py and processing the data to get a complete records_final.pkl

  vim processing.py
  
  # line 364-366
  # med_file = './input/PRESCRIPTIONS.csv'
  # diag_file = './input/DIAGNOSES_ICD.csv'
  # procedure_file = './input/PROCEDURES_ICD.csv'
  
  python processing.py

• run ddi_mask_H.py to get the ddi_mask_H.pkl

  python ddi_mask_H.py

Step 3: run the code, we use

python HKRec.py

here is the argument:

usage: HKRec.py [-h] [--Test] [--model_name MODEL_NAME]
               [--resume_path RESUME_PATH] [--lr LR]
               [--target_ddi TARGET_DDI] [--kp KP] [--dim DIM]

optional arguments:
  -h, --help            show this help message and exit
  --Test                test mode
  --model_name MODEL_NAME
                        model name
  --resume_path RESUME_PATH
                        resume path
  --lr LR               learning rate
  --batch_size          batch size 
  --emb_dim             dimension size of embedding
  --max_len             max number of recommended medications
  --beam_size           number of ways in beam search

If you cannot run the code on GPU, just change line 117, "cuda" to "cpu".

Please feel free to contact me 2021212134@nwnu.edu.cn for any question.