OmniGenBench, curated

OmniGenBench, as published in quality-curated genomic benchmarks - one format, fixed row order, a permanent ID on every row. 21 datasets, 44 files, 175,725 rows, one gzipped CSV per split.

Getting the data

Two packages are the way in: genomic-benchmarks-data for people, genomic-benchmarks-data4agents for agents, the same functions either way. They resolve the URL, check the checksum, and carry each dataset's QC results, which this repository does not.

pip install genomic-benchmarks-data

from genomic_benchmarks_data import load_dataset

df = load_dataset("OmniGenBench", "lncrna_g_max", "train")

A number measured on these files belongs on the leaderboard, which is what makes two of them comparable, and which prints what QC found about a dataset above every table.

• genomic-benchmarks.github.io - the project: the QC tool, the curated corpus and the leaderboard.
• github.com/genomic-benchmarks - all three as source, and issues.

Datasets

dataset	task	species	train	val	test
lncrna_g_max	Binary classification	Glycine max	8000	-	490
lncrna_m_esculenta	Binary classification	Manihot esculenta	4934	-	360
lncrna_s_bicolor	Binary classification	Sorghum bicolor	8000	-	734
lncrna_s_lycopersicum	Binary classification	Solanum lycopersicum	7274	-	1000
lncrna_t_aestivum	Binary classification	Triticum aestivum	8000	-	1000
lncrna_z_mays	Binary classification	Zea mays	8000	-	1000
poly_a_arabidopsis_thaliana	Binary classification	Arabidopsis thaliana	8000	-	1000
poly_a_chlamydomonas_reinhardtii	Binary classification	Chlamydomonas reinhardtii	8000	-	1000
poly_a_medicago_truncatula	Binary classification	Medicago truncatula	8000	-	1000
poly_a_oryza_sativa_indica_group	Binary classification	Oryza sativa	8000	-	1000
poly_a_oryza_sativa_japonica_group	Binary classification	Oryza sativa	8000	-	1000
poly_a_trifolium_pratense	Binary classification	Trifolium pratense	8000	-	1000
pro_seq_m_esculenta	Binary classification	Manihot esculenta	8000	-	812
promoter_strength_leaf	Regression	Nicotiana tabacum	8000	-	1000
promoter_strength_protoplast	Regression	Zea mays	8000	-	1000
splicing_arabidopsis_thaliana_acceptor	Binary classification	Arabidopsis thaliana	8000	-	1000
splicing_arabidopsis_thaliana_donor	Binary classification	Arabidopsis thaliana	8000	-	1000
te_prediction_arabidopsis	Binary classification	Arabidopsis thaliana	3399	424	426
te_prediction_rice	Binary classification	Oryza sativa	4697	587	588
terminator_strength_leaf	Regression	Nicotiana tabacum	8000	-	1000
terminator_strength_protoplast	Regression	Zea mays	8000	-	1000

Files are v1/<dataset>/<dataset>_<split>.csv.gz, where v1 is the version of the data. A dataset whose sequences change appears under a new prefix rather than being rewritten here.

Licence and provenance

• Collection DOI: https://doi.org/10.48550/arXiv.2505.14402
• Covers every dataset here: MIT (https://huggingface.co/spaces/yangheng/OmniGenBenchLeaderboard/blob/main/README.md)

Every row has an identity

Each file carries two columns the upstream release does not:

• id - <collection>:<dataset>:<version>:<split>:<row index>, the index zero-padded to 7 digits.
• row_index - the row's 0-based position in this file.

Row order is fixed within a major version, so an id names the same sequence for as long as that version exists - which is what makes a reported number traceable to what it was computed on. The corpus metadata carries the SHA-256 of every split file, so a copy can be checked against the audited one.

Citation

Cite the collection's own publication (https://doi.org/10.48550/arXiv.2505.14402) for the data, and Automated quality control for genomic sequence benchmarks reveals pervasive single-feature separability and train-test leakage for the curation and the audit.