# Random Promoter DREAM Challenge 2022

## Facts

- Provider: DREAM Challenges organizers (Random Promoter DREAM Challenge 2022); packaged by Hugging Face Bio
- License: CC-BY-4.0 (https://zenodo.org/api/records/10633252)
- Size: 10 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/HuggingFaceBio/random-promoter-dream-2022
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.10633252
- Category: Genomics & Bioinformatics
- Source: https://doi.org/10.5281/zenodo.10633252
- Page: https://scigantic.com/archive/383719ff-331a-53a7-8278-e1581a33f60c
- Keywords: Genomics & Bioinformatics, benchmark, Parquet, Hugging Face

## About

Hugging Face packaging of the processed files from the Random Promoter DREAM Challenge 2022 (Zenodo record 10633252), a community benchmark for predicting gene expression from DNA sequence. Participants were given expression measurements for millions of randomly generated promoter sequences assayed in yeast and asked to train models that predict promoter activity, then scored on a separate set of designed test promoters. The supervised config has train, validation and labeled test splits with columns sequence, activity (measured promoter activity), sequence_length, source_file and row_id. Additional configs hold the unlabeled challenge test sequences, test subset membership IDs (the subsets used for scoring) and the public leaderboard IDs, each traced back to the archive member and line it came from. Measured 2026-11-14: the supervised split is four parquet files (two train shards of about 155 MB each, a 34 MB validation file and a small test file), about 0.35 GB, with repository storage of about 0.4 GB. Load with datasets.load_dataset('HuggingFaceBio/random-promoter-dream-2022', 'supervised') or pandas.read_parquet. Typical uses are sequence-to-function modelling, massively parallel reporter assay style regression, and benchmarking DNA language models and convolutional or transformer architectures for cis-regulatory grammar. The source Zenodo record is licensed CC BY 4.0; cite the DREAM Challenge data release.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("HuggingFaceBio/random-promoter-dream-2022")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `challenge_test_sequences/train-00000-of-00001.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/383719ff-331a-53a7-8278-e1581a33f60c
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/383719ff-331a-53a7-8278-e1581a33f60c
- Over MCP: `get_schema_card` with id `383719ff-331a-53a7-8278-e1581a33f60c` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Genomics & Bioinformatics datasets](https://scigantic.com/topics/genomics-and-bioinformatics)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
