license:
- cc-by-4.0
- other
pretty_name: mRNABench (curated)
size_categories:
- 1M<n<10M
tags:
- genomics
- benchmark
- rna
- regression
mRNABench, curated
mRNABench, as published in quality-curated genomic benchmarks - one format, fixed row order, a permanent ID on every row. 11 datasets, 33 files, 1,413,237 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("mRNABench", "mrl-sample-designed", "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 |
|---|---|---|---|---|---|
mrl-sample-designed |
Regression | Homo sapiens | 70011 | 15003 | 15003 |
mrl-sample-egfp-m1pseudo |
Regression | Homo sapiens | 222927 | 47770 | 47771 |
mrl-sample-egfp-pseudo |
Regression | Homo sapiens | 222927 | 47770 | 47771 |
mrl-sample-egfp-unmod |
Regression | Homo sapiens | 222927 | 47770 | 47771 |
mrl-sample-mcherry |
Regression | Homo sapiens | 134257 | 28770 | 28770 |
mrl-sample-varying |
Regression | Homo sapiens | 74571 | 15979 | 15980 |
mrl-sugimoto |
Regression | Homo sapiens | 8186 | 1753 | 1754 |
rnahl-human |
Regression | Homo sapiens | 9057 | 1942 | 1940 |
rnahl-mouse |
Regression | Mus musculus | 9606 | 2058 | 2059 |
translation-efficiency-human |
Regression | Homo sapiens | 7196 | 1543 | 1542 |
translation-efficiency-mouse |
Regression | Mus musculus | 7597 | 1628 | 1628 |
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/10.1101/2025.07.05.662870
| dataset | licence |
|---|---|
mrl-sample-designed |
No restrictions (GEO deposition) |
mrl-sample-egfp-m1pseudo |
No restrictions (GEO deposition) |
mrl-sample-egfp-pseudo |
No restrictions (GEO deposition) |
mrl-sample-egfp-unmod |
No restrictions (GEO deposition) |
mrl-sample-mcherry |
No restrictions (GEO deposition) |
mrl-sample-varying |
No restrictions (GEO deposition) |
mrl-sugimoto |
CC BY 4.0 |
rnahl-human |
CC BY 4.0 |
rnahl-mouse |
CC BY 4.0 |
translation-efficiency-human |
No restrictions (author statement) |
translation-efficiency-mouse |
No restrictions (author statement) |
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/10.1101/2025.07.05.662870) 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.