README.md

---
pretty_name: DDInter Inductive Reasoning
license: apache-2.0
annotations_creators:
- no-annotation
language:
- en
task_categories:
- question-answering
- text-generation
---

## 🔗 This dataset is part of the study:  
**_K-Paths: Reasoning over Graph Paths for Drug Repurposing and Drug Interaction Prediction_**

- 📖 [Read the Paper](https://arxiv.org/abs/2502.13344)
- 💾 [GitHub Repository](https://github.com/rsinghlab/K-Paths)


# DDInter: Inductive Reasoning Dataset

DDInter provides drug–drug interaction (DDI) data labeled with three severity levels (`Major`, `Moderate`, `Minor`). 
Each entry includes two drugs, an interaction label, drug descriptions, and structured/natural language representations of multi-hop reasoning paths between them.
The multi-hop paths were extracted from a biomedical knowledge graph (Hetionet) using K-Paths.

This dataset is useful for evaluating:
- Path-based biomedical reasoning
- Knowledge graph inference
- Inductive reasoning



## 💡 Columns

| Column              | Description                                      |
|---------------------|--------------------------------------------------|
| `drug1_db`, `drug2_db` | DrugBank IDs                                 |
| `drug1_id`, `drug2_id` | Node IDs from the KG                         |
| `drug_pair`         | Optional drug pair identifier                    |
| `drug1_name`, `drug2_name` | Names of each drug                      |
| `drug1_desc`, `drug2_desc` | Descriptions of each drug               |
| `label`             | Interaction severity (e.g., `Moderate`)          |
| `label_idx`         | Numeric version of the label                     |
| `all_paths`, `all_paths_str`, `path_str` | KG paths for reasoning     |

## 📦 Example Record

```json
{
  "drug1_name": "Mannitol",
  "drug2_name": "Castor oil",
  "label": "Moderate",
  "path_str": "Mannitol (Compound) binds ABCB1 (Gene) and ABCG2 (Gene)..."
}

## 📥 How to Load with pandas

```python
import pandas as pd

splits = {
    "train": "ddinter/train.json",
    "test": "ddinter/test.json"
}

train = pd.read_json("hf://datasets/Tassy24/K-Paths-inductive-reasoning-ddinter/" + splits["train"], lines=True)
test = pd.read_json("hf://datasets/Tassy24/K-Paths-inductive-reasoning-ddinter/" + splits["test"], lines=True)



### 📜 Citation (BibTeX)

```bibtex
@article{abdullahi2025k,
  title={K-Paths: Reasoning over Graph Paths for Drug Repurposing and Drug Interaction Prediction},
  author={Abdullahi, Tassallah and Gemou, Ioanna and Nayak, Nihal V and Murtaza, Ghulam and Bach, Stephen H and Eickhoff, Carsten and Singh, Ritambhara},
  journal={arXiv preprint arXiv:2502.13344},
  year={2025}
}