README.md

---
license: gpl-3.0
tags:
- biology
pretty_name: ProtHGT Knowledge Graph Data & Pretrained Checkpoints
---

# ProtHGT Knowledge Graph Data & Pretrained Checkpoints
This repository provides the **knowledge graph (KG) `.pt` files** and **pretrained model checkpoints** used in **ProtHGT: Heterogeneous Graph Transformers for Automated Protein Function Prediction Using Biological Knowledge Graphs and Language Models**.
- **Code (training & prediction)**: https://github.com/HUBioDataLab/ProtHGT

---

## What’s Inside

### data/
PyTorch Geometric-compatible KG files:
- Full KG file (e.g., `prothgt-kg.pt`)
- Train/validation/test splits (e.g., `prothgt-*-graph.pt`)
- Alternative KG versions under `alternative_protein_embeddings/` (e.g., `esm2/`, `prott5/`), where the protein node features differ by embedding type.

**Available Files**
```
├── prothgt-kg.pt                      # The default full knowledge graph containing TAPE embeddings as the initial protein representations.
├── prothgt-train-graph.pt             # Training set (80% of the default full KG).
├── prothgt-val-graph.pt               # Validation set (10% of the default full KG).
├── prothgt-test-graph.pt              # Test set (10% of the default full KG).
└── alternative_protein_embeddings/    # Contains alternative KGs with different protein representations.
    ├──apaac/
    │   └── ...
    ├──esm2/
    │   └── ...
    └──prott5/
        └── ...
```


### models/
Pretrained ProtHGT models (`.pt`). Models are provided:
- per GO sub-ontology (e.g., Molecular Function / Biological Process / Cellular Component)
- per protein embedding type (default vs `esm2` / `prott5` / etc.)

**Important:** Use a model checkpoint that matches the KG embedding variant you are using.

**Available Files**
```
├── prothgt-model-molecular-function.pt      # Pretrained ProtHGT checkpoint for Molecular Function (default/TAPE-based KG).
├── prothgt-model-biological-process.pt      # Pretrained ProtHGT checkpoint for Biological Process (default/TAPE-based KG).
├── prothgt-model-cellular-component.pt      # Pretrained ProtHGT checkpoint for Cellular Component (default/TAPE-based KG).
└── alternative_protein_embeddings/          # Models trained with alternative protein representations.
    ├── esm2/
    │   └── ...
    └── prott5/
        └── ...
```

---

### How to Use (Training & Prediction)
To train or run inference, follow the instructions in the GitHub repository: https://github.com/HUBioDataLab/ProtHGT

Key scripts:
- `train.py` — trains ProtHGT using the provided KG splits
- `predict.py` — runs inference using pretrained checkpoints

---

### Citation
Please refer to our preprint for more information. If you use the ProtHGT method or the datasets provided in this repository, please cite this paper:
Ulusoy, E., & Dogan, T. (2025). ProtHGT: Heterogeneous Graph Transformers for Automated Protein Function Prediction Using Biological Knowledge Graphs and Language Models (p. 2025.04.19.649272). bioRxiv. [Link](https://doi.org/10.1101/2025.04.19.649272)

---

### Licensing
Copyright (C) 2025 HUBioDataLab

This dataset is released under GPL-3.0.