README.md

---
license: cc-by-nc-4.0
dataset_info:
  features:
    - name: accession
      dtype: string
    - name: name
      dtype: string
    - name: Full Name
      dtype: string
    - name: taxon
      dtype: string
    - name: sequence
      dtype: string
    - name: function
      dtype: string
    - name: AlphaFoldDB
      dtype: string
  splits:
    - name: train
      num_bytes: 168911398
      num_examples: 248315
    - name: test
      num_bytes: 3470418
      num_examples: 4203
    - name: validation
      num_bytes: 3443875
      num_examples: 4172
  download_size: 134650306
  dataset_size: 175825691
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
      - split: validation
        path: data/validation-*
---



# Dataset Card for Prot2Text-Data

**Page:** [Prot2Text](http://nlp.polytechnique.fr/prot2text#proteins) <br>
**Paper:** [https://arxiv.org/abs/2307.14367](https://arxiv.org/abs/2307.14367) <br>
**Github:** [https://github.com/hadi-abdine/Prot2Text](https://github.com/hadi-abdine/Prot2Text) <br>
**Authors:** Hadi Abdine<sup>(1)</sup>, Michail Chatzianastasis<sup>(1)</sup>, Costas Bouyioukos<sup>(2, 3)</sup>, Michalis Vazirgiannis<sup>(1)</sup><br>
<sup>**(1)**</sup>DaSciM, LIX, École Polytechnique, Institut Polytechnique de Paris, France.<br>
<sup>**(2)**</sup>Epigenetics and Cell Fate, CNRS UMR7216, Université Paris Cité, Paris, France.<br>
<sup>**(3)**</sup>Bioinformatics Research Laboratory, Department of Biological Sciences, University of Cyprus, Nicosia, Cyprus.<br>

**Prot2Text** paper is published in **AAAI 2024**. Preliminary versions of the paper were accepted as a spotlight at [DGM4H@NeurIPS 2023](https://sites.google.com/ethz.ch/dgm4h-neurips2023/home?authuser=0) and [AI4Science@NeurIPS 2023](https://ai4sciencecommunity.github.io/neurips23.html). 

## Dataset Description

This dataset is designed for training the [Prot2Text](https://huggingface.co/habdine/Prot2Text-Base-v1-0) framework. It  contains information for 256,690 proteins and is comprised of three modalities:

* Protein sequence (amino acid sequence).
* Protein structure using its AlphaFold accession ID.
* Textual description of the protein

The dataset is built from the SwissProt database, a component of UniProtKB Release 2022_04.



## Dataset Structure

#### Data Fields:

* **name:** (string) codename of the protein.
* **Full Name:** (string) Full name of the protein.
* **sequence:** (string) Amino acid sequence of the protein.
* **AlphaFoldDB:** (string) Accession ID of the protein in AlphaFold database.
* **taxon:** (string) Species information for the protein.
* **function:** (string) Textual description of the protein.


#### Data Splits:

The dataset is split into training, validation, and test sets with a maximum sequence similarity threshold of 40% within each set using the CD-HIT clustering algorithm.

* **Train:** 248,315 proteins
* **Validation:** 4,172 proteins
* **Test:** 4,203 proteins

### Considerations for Using the Data

The dataset is built from a single source (SwissProt). Consider incorporating data from other sources to increase diversity.
The textual descriptions may contain biases present in the original database. Be mindful of these biases when using the data for downstream tasks.


### License
We are releasing this dataset under the terms of [CC-BY-NC-4.0](https://creativecommons.org/licenses/by-nc/4.0/deed.en).


### Citation
Please cite this dataset and the original sources if you use this dataset in your work
```
@inproceedings{abdine2024prot2text,
  title={Prot2Text: Multimodal Protein's Function Generation with GNNs and Transformers},
  author={Abdine, Hadi and Chatzianastasis, Michail and Bouyioukos, Costas and Vazirgiannis, Michalis},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  pages={10757--10765},
  year={2024}
}
```