Skip to content

Repository files navigation

TopoEdit-Off

DOI

TopoEdit-Off is a source-aware benchmark resource for high-mismatch genome-editing off-target prioritization.

This repository contains the public software/resource release accompanying the manuscript:

TopoEdit-Off: a source-aware benchmark for high-mismatch genome-editing off-target prioritization

What this repository provides

TopoEdit-Off organizes public genome-editing off-target evidence into a compact, source-aware benchmark. The resource is designed to support:

  • comparison of conventional nuclease, TOPO-seq, and base-editor observed-positive spaces;
  • mismatch-regime analysis across evidence sources;
  • size-matched top-K ranking evaluation;
  • validation-priority panel design;
  • reproducible access to compact benchmark summaries and release metadata.

This repository is intentionally minimal. It contains public-facing summary tables, ranking summaries, validation-priority panels, manifests, checksums, and a minimal smoke test. Large candidate-level derived tables are not stored in this GitHub repository.

Key benchmark numbers

Item Value
Guide panel 100 guides
Generated candidate universe 1,961,227 rows
Source-aware known-positive rows 25,457 rows
Mismatch-6 candidate rows 1,637,800 rows
TOPO source-positive guide support 6 guides
Source-aware base-editor-positive guide support 11 guides
Core validation-priority panel 96 candidates

Repository structure

data/benchmark/          Core source-aware benchmark summaries
data/ranking/            Size-matched top-K ranking evaluation
data/validation_panels/  Validation-priority candidate panels
demo/                    Minimal smoke test
metadata/                File manifest, checksums, and data dictionary
tools/                   Lightweight release validator

Main public tables

Path Purpose
data/benchmark/source_aware_label_inventory.tsv Summary of source-aware label counts
data/benchmark/mismatch_bucket_coverage.tsv Mismatch-regime coverage by label subset
data/benchmark/candidate_burden_by_family.tsv Candidate burden by evidence family
data/ranking/size_matched_topk_recovery_summary.tsv Size-matched observed-positive recovery at top K
data/ranking/delta_vs_sequence_baseline_summary.tsv Change in recovery relative to the sequence baseline
data/ranking/rank_methods.tsv Ranking methods used in the public summary
data/ranking/metric_definitions.tsv Metric definitions and interpretation notes
data/validation_panels/validation_panel_core96.tsv Core validation-priority panel
data/validation_panels/panel_tier_summary.tsv Core, expanded, and backup panel sizes

Minimal smoke test

The demo checks that the public release is readable and internally consistent. It does not rerun the full candidate-generation or model-training pipeline.

python -m pip install pandas
python demo/run_minimal_demo.py
cat demo/output/minimal_demo_summary.md

Expected generated files:

demo/output/minimal_demo_summary.md
demo/output/top1000_recovery_summary.tsv

Run the repository validator:

python tools/validate_release.py --root .

Expected result:

[PASS] minimal TopoEdit-Off public release validation passed.

Interpretation policy

TopoEdit-Off reports observed-positive recovery and validation-priority behavior.

The resource does not claim that:

  • unlabeled candidates are experimentally validated negatives;
  • ranking scores are calibrated off-target probabilities;
  • validation-priority panels are wet-validated panels;
  • TOPO-seq or base-editor off-target biology is universally high-mismatch in every setting.

The intended use is benchmark analysis, source-aware comparison, and prioritization of candidate panels for future experimental validation.

What is not included

This GitHub repository does not include raw SRA data, genome files, alignment outputs, BAM files, FASTQ files, or large candidate-level derived tables.

Large derived tables should be archived separately as Zenodo Dataset records when they are cited directly in the manuscript.

Citation

Cite the archived Zenodo release: DOI 10.5281/zenodo.20154047. See CITATION.cff for citation metadata.

License

MIT.

Version

v1.1.1

About

Source-aware benchmark and validation-priority resource for high-mismatch genome-editing off-target prioritization.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages