DrugBankNER

ETL pipeline for processing DrugBank data and aligning entities and identifiers with RTX-KG2 concepts. The resulting output is intended for use as training or evaluation data in downstream knowledge graph tasks.

---

Prep Work

1. Download DrugBank XML

1. Create an account at https://go.drugbank.com
2. Run:

   ./download_data.sh

   This will place the DrugBank XML file in the data/ directory.

---

2. Obtain the RTX / ARAX Node Synonymizer

You will need a compatible node synonymizer SQLite database for the KG version you are using.

• Recommended: Ask a team member for a local copy of the node synonymizer database
  (this is useful if you do not have access to the RTX database server).

• Alternative: Request access to the RTX database server (arax-databases.rtx.ai).
  If access is granted, the scripts will automatically download the appropriate node synonymizer database when it is not found locally.

⚠️ The node synonymizer version must match the KG version passed via --kg-version.
The scripts first check for a local database and only attempt a download if it is not already available.

---

3. Environment Setup

Create and activate a conda environment:

conda create --name drugbank_ner python=3.11.10
conda activate drugbank_ner

Install required packages:

pip install xmltodict==0.14.2
pip install pandas==2.2.3
pip install spacy==3.8.2
pip install scispacy==0.5.5

Optional: GPU support

Check your CUDA version:

nvidia-smi

Then install the matching CuPy package:

pip install cupy-cuda<your_cuda_version>x

---

4. Install ScispaCy Models

pip install https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/releases/v0.5.3/en_core_sci_lg-0.5.3.tar.gz
pip install https://s3-us-west-2.amazonaws.com/ai2-s2-scispacy/releases/v0.5.3/en_core_sci_scibert-0.5.3.tar.gz

---

Running the Tool

Important
As of the latest refactor, all scripts require the --kg-version argument.
This ensures that downloaded databases and alignment logic are consistent with the intended knowledge graph version.

---

1. Perform Named Entity Recognition

Run perform_NER.py to perform named entity recognition and concept alignment on DrugBank text fields.

python perform_NER.py --kg-version 2.10.2

Optional arguments

--db-host        Database file host (default: arax-databases.rtx.ai)
--db-username    Database file username (default: rtxconfig)
--db-port        Database file port (default: 22)
--ssh-key        Path to SSH private key (optional; uses SSH agent if omitted)
--ssh-password   SSH password (optional; prefer key or agent; can also set SSH_PASSWORD env var)
--out-dir        Output directory for downloaded database files (default: ./data)

Example:

python perform_NER.py \
  --kg-version 2.10.2 \
  --out-dir ./data \
  --ssh-key ~/.ssh/id_rsa

---

2. Extract and Align Identifiers

Run look_for_identifiers.py to extract, synonymize, and align DrugBank identifiers with RTX-KG2.

python look_for_identifiers.py --kg-version 2.10.2

This script supports the same optional connection and output arguments as perform_NER.py.

---

Output

After successfully running both scripts, the final aligned output will be written to:

./data/DrugBank_aligned_with_KG2.json

---

Notes

• The --kg-version argument must follow the format X.Y.Z (e.g. 2.10.2)
• Ensure all downloaded database artifacts correspond to the specified KG version
• SSH key–based authentication is strongly recommended over password-based access