Benchmarks/Genomic Benchmarks

suite · audited · verified 2026-07-22

Genomic Benchmarks

A versioned collection of nine DNA sequence-classification datasets covering regulatory elements, promoters, enhancers, open chromatin, species, and coding-context discrimination.

Nucleic-acid result atlas: B06 results (same_entity)

Benchmark definition

What is counted

Version: package-1.0.0-snapshot
Total: 9 (benchmark dataset classification tasks)
Task formats: binary genomic sequence classification; multiclass genomic sequence classification
Capabilities: PredictionClassification
Modalities: DNA or RNA sequence

Version history

Version	Status	Release / as-of	Total	Formal tracks
package-1.0.0-snapshot `genomic-benchmarks-package-1-0-0-snapshot`	rolling	2023-05-01	9 (benchmark dataset classification tasks)	None registered

Tracks and subsets

ID	Count	Basis	Partition?	Notes
Core regulatory-element datasets `genomic-benchmarks-core-regulatory`	6	benchmark dataset classification tasks	Exclusive & exhaustive	Drosophila enhancer plus five human regulatory datasets.
Demonstration datasets `genomic-benchmarks-demo`	2	benchmark dataset classification tasks	Exclusive & exhaustive	Coding-versus-intergenic and human-versus-worm demos.
Dummy dataset `genomic-benchmarks-dummy`	1	benchmark dataset classification tasks	Exclusive & exhaustive	Small mouse-enhancer prototyping dataset.

Scientific Task Atlas

Scientific task classification

complete for package-1.0.0-snapshot · as of 2026-07-22. Each of the nine version-0 dataset directories is assigned once using the creator paper's Table 1 labels.

Scientific task	Coverage	Count	Mapping	Evidence
Enhancer activity prediction	explicitly-in-scope	4 tasks Benchmark dataset classification tasks.	official-taxonomy high confidence	`genomic-benchmarks-paper-definition-evidence` Mouse enhancer, Drosophila enhancer, human enhancer Cohn, and human enhancer Ensembl datasets.
Promoter detection	explicitly-in-scope	1 tasks Benchmark dataset classification tasks.	official-taxonomy high confidence	`genomic-benchmarks-paper-definition-evidence` Human non-TATA promoter classification.
Epigenetic-mark prediction	explicitly-in-scope	1 tasks Benchmark dataset classification tasks.	official-taxonomy high confidence	`genomic-benchmarks-paper-definition-evidence` Human open-chromatin region classification.
DNA sequence analysis	explicitly-in-scope	3 tasks Benchmark dataset classification tasks.	official-taxonomy high confidence	`genomic-benchmarks-paper-definition-evidence` Two demo coding-versus-intergenic datasets and the human regulatory-region multiclass dataset.

Scientific coverage notes

Domain	Coverage	Count	Interpretation
Genomics	explicitly-in-scope	9	All nine formal entries are genomic DNA sequence-classification datasets.
Epigenomics	explicitly-in-scope	1	The human open-chromatin-region dataset directly tests chromatin accessibility classification.

Evaluation registry

Works and run settings

A setting change—scope, prompt, tools, budget, grader, or repeats—creates a separate run. Charts never cross a comparability group.

Evaluation run

genomic-benchmarks-creator-full

From Genomic benchmarks: a collection of datasets for genomic sequence classification

Scopefull · n=9

ShotsNot applicable

TurnsNot applicable

System prompt publicNot applicable

Reasoning / effortNot applicable

BrowserNot applicable

InternetNot applicable

DatabasesNot applicable

Code executionNot applicable

ContainerNot reported

External toolsTensorFlow and PyTorch CNN baseline workflows

Token budgetNot applicable

Time / cost budgetNot reported

TemperatureNot applicable

SeedNot reported

RepeatsNot reported

Graderdeterministic classification scorer · human review: no

StatisticsAccuracy and F1 are reported independently for each dataset and framework.

ContaminationDataset-specific train/test splits; duplicate and background-generation controls follow each construction notebook.

Metrics, results, and full protocol

Metrics

Metric	Kind / baseline	Unit	Aggregation	Threshold / tolerance
Accuracy	absolute	percent	held-out test sequences per dataset	Not reported
F1 score	absolute	percent	held-out test sequences per dataset	Not reported

No numeric result rows are published yet; the verified protocol remains useful.

Evidence

table: Methods and Tables 1-2 (Lists all nine datasets, train/test design, CNN workflows, and accuracy/F1 results.) — supports /scope, /protocol, /metrics

Evidence and change history

Source locators remain visible; expand an item to inspect the exact Registry fields it supports.

Genomic benchmarks: a collection of datasets for genomic sequence classification · table: Construction and content, Overview of Datasets, Tables 1-2 (Lists nine datasets, sequence counts, classes, categories, train/test design, and creator CNN metrics.) · Supports 26 fields

Open source →

/name
/aliases
/summary
/kind
/organizations
/release_date
/latest_version
/domains
/capabilities
/modalities
/task_formats
/task_counts/total
/task_counts/basis
/task_counts/subsets
/coverage_notes
/access/level
/access/license
/resources
/versions/0/release_date
/versions/0/task_counts/total
/versions/0/task_counts/basis
/versions/0/task_counts/subsets
/scientific_task_classification/entries/0
/scientific_task_classification/entries/1
/scientific_task_classification/entries/2
/scientific_task_classification/entries/3

genomic-benchmarks-repository-resource · repository-path: README.md; setup.py; datasets/*/metadata.yaml; experiments/README.md at 605d8539830e16c85abe7826990958303ffc5e1c (Confirms package version, nine dataset directories, version metadata, public loaders, licenses, and baseline metrics.) · Supports 9 fields

Open source →

/latest_version
/access/tasks
/access/artifacts
/access/grader
/access/license
/resources
/implementations
/versions/0/as_of
/versions/0/formal_tracks

View source-level modification history on GitHub →

Access & reproducibility

Access: Fully open
License: Apache-2.0 for package code; reference genomes and adapted source datasets retain source-specific terms.
Grader: Deterministic classification accuracy and F1 calculations on held-out test sets.
Implementations: 1 recorded · official