# scELMo Cell Type Embeddings

This directory contains cell type embeddings generated using scELMo methodology.

## Source

These embeddings are converted from the official scELMo repository:
**https://github.com/HelloWorldLTY/scELMo**

## Model Information

- **Model**: gpt-3.5-turbo
- **Embedding Dimension**: 1536
- **Type**: celltype
- **Aggregation Mode**: wa
- **API Model**: text-embedding-ada-002

## Files

- `gene_embeddings.pkl`: Gene embeddings dictionary in PerturbLab format
  - Format: `{'embeddings': {gene_name: embedding_array}, 'gene_list': [gene_names]}`
- `config.json`: Model configuration

## Usage

```python
from perturblab.model.scelmo import scELMoModel

# Load model
model = scELMoModel.from_pretrained('scelmo-celltypes-gpt-3.5')

# Use embeddings
embeddings = model.predict_embeddings(adata, aggregation_mode='wa')
```

## Citation

If you use these embeddings, please cite the original scELMo paper:

```bibtex
@article{liu2023scelmo,
  title={scELMo: Embeddings from Language Models are Good Learners for Single-cell Data Analysis},
  author={Liu, Tianyu and Chen, Tianqi and Zheng, Wangjie and Luo, Xiao and Zhao, Hongyu},
  journal={Cell Patterns (in press)},
  pages={2023--12},
  year={2025},
  publisher={Cell Press}
}
```