CoolFace
Datasetpublic

marin-dna/zoonomia-v1-v3_ccre_non_promoter-tutorial

Tiny Zoonomia non-promoter cCRE tutorial sample This dataset contains 256 unchanged rows from the public marin-dna/zoonomia-v1-v3_ccre_non_promoter cross-mammal enhancer dataset. It exists only to keep MarinDNA's local CPU training tutorial fast; it is not an independent biological dataset or a representative benchmark. Contents data/train.jsonl contains the first 256 records from source shard data/train/shard_0000.jsonl.zst at immutable source revision… See the full description on the dataset page: https://huggingface.co/datasets/marin-dna/zoonomia-v1-v3_ccre_non_promoter-tutorial.

sourceHugging Faceupdated 2mo agoView on Hugging Face
0likes10downloads
Dataset Card

Tiny Zoonomia non-promoter cCRE tutorial sample

This dataset contains 256 unchanged rows from the public `marin-dna/zoonomia-v1-v3_ccre_non_promoter` cross-mammal enhancer dataset. It exists only to keep MarinDNA's local CPU training tutorial fast; it is not an independent biological dataset or a representative benchmark.

Contents

data/train.jsonl contains the first 256 records from source shard data/train/shard_0000.jsonl.zst at immutable source revision 862485aa18eed53a53e693ba4c2eb45e0afc5087. The source dataset was globally shuffled with seed 42 before sharding.

Every record retains the source schema:

FieldDescription
query_nameHuman anchor identifier.
speciesZoonomia target species.
t_chromTarget chromosome.
t_start, t_end0-based, half-open target interval; always 255 bp.
t_strandProjection strand (+ or -).
t_src_sizeTarget chromosome size.
sequenceStrand-aware 255 bp DNA sequence.
augmentationOriginal (+) or reverse-complement (-) augmentation.

Provenance

The deterministic builder copies and validates the rows without transforming their contents:

`scripts/issue412_build_tiny_enhancer_dataset.py`

The source dataset card documents the ENCODE cCRE V4 non-promoter labels, Zoonomia projection, reverse-complement augmentation, and upstream pipeline.