Benchmark_NP

Reproducible analysis code for the nanopore basecaller benchmark.

Reproduce the analysis

git clone https://github.com/AlexanderM-M/Benchmark_NP.git
cd Benchmark_NP
conda env create -f environment.yml
conda activate benchmark_env
python run_analysis.py

Before the final command, place input_excel.xlsx beside run_analysis.py. The script validates the workbook, writes tables to results/, and writes publication figures to figures/.

Input data

The workbook must contain three sheets:

• raw_fast
• raw_hac
• raw_sup

Each sheet uses two header rows. Required columns, units, missing-value conventions, and validation rules are listed in DATA_DICTIONARY.md.

The study workbook is not included because it may contain restricted patient-level information. Add the data-access location and conditions here when they are finalized. The pipeline records a SHA-256 checksum in results/analysis_manifest.json, allowing collaborators to confirm that they used the same workbook without publishing it.

Outputs

results/ contains:

• Classification, transition, concordance, QC, operating-range, and workflow summary tables
• Tissue-only histology concordance workbooks
• The fitted operating-model summary
• An analysis manifest containing software versions and the input checksum

figures/ contains:

• Figures 1–5 in PNG, PDF, and SVG formats
• The categorical sample heatmap in PNG, PDF, and SVG formats
• Figure-specific statistical tables and the heatmap source table

Both directories are generated and ignored by Git because some outputs can contain sample identifiers. Review generated files before sharing them.

Notebook

Benchmark_analysis_clean.ipynb is the annotated analysis notebook used by run_analysis.py. Its committed outputs and execution counts are intentionally cleared. Users do not need to open or execute it manually.

Verification

Run the repository checks with:

python -m unittest discover -s tests -v

These checks verify Python syntax, notebook privacy, expected repository files, and workbook naming conventions. GitHub Actions runs the same checks on every push.

Citation

Citation metadata are provided in CITATION.cff. If you use this analysis, cite the accompanying article and the archived software release once its DOI is available.

License

This project is freely available under the MIT License.