Data and reproducibility

Documentation · Benchmarks

Tutorial data

All tutorial inputs are simulated.

The repository contains 24 reference and 48 control FASTA sequences generated with seed 42, a script to regenerate them, and a demonstration that creates a fictional RBP score matrix with seed 123.

No biological case-study data are needed.

python examples/run_demo.py --out results/demo

This small API exercise is not the study's in silico benchmark.

Benchmark data

benchmarks/reported_summary.tsv contains approximate summaries from the manuscript and explicit eCLIP/PEKA counts from the presentation.

Original per-dataset measurements and benchmark pipelines are not included in the supplied material.

The following resources still need to be deposited for full reproduction:

Resource	Required content
Synthetic sequences	FASTA groups, injected motifs, interval ground truth and seeds
Comparator workflows	Scripts, commands, exact environments and logs
Performance tables	Per-dataset scores with valid, failed, missing and undefined status
Transcript resources	Exact sequence files, annotation and selection rules
eCLIP resources	ENCODE file accessions, transcript groups, peak and control intervals
RBP score matrices	PEKA source sheets, profiles, preprocessing and checksums
Analysis notebooks	Complete evaluation and figure-generation code

The PEKA resource cited in the benchmark methods is Additional file 5, Table S4 of Kuret et al. (2022).

The exact curated matrix used in the evaluation is not bundled.

The repository provides a reserved PEKA folder with matrix-format instructions.

Run records

For each analysis record:

• Code commit or release.
• Python and package versions.
• Platform.
• Input checksums and annotation release.
• Group membership.
• k-mer length.
• Statistical test and correction.
• Selection thresholds.
• Random seeds.

Preserve complete result tables as well as filtered signatures.

Capture installed versions with:

python -m pip freeze > environment.txt

The snapshot fixes the final-window counting error and stabilizes K-RBP background ordering. These changes can affect results; benchmark summaries have not been recomputed.

See CHANGELOG.

Provenance

provenance/source-manifest.json records checksums of the supplied code and documentation inputs.

provenance/source-changes.patch records code edits.

The draft manuscript and complete working presentation will be distributed in the repository.