Oral annonacin pre-QIVIVE analysis

This repository accompanies the Research Brief “What Sparse Brain Kinetic Data Can—and Cannot—Tell Us About Oral Annonacin.” It reproduces the deterministic calculations that relate a published rat whole-brain measurement 48 h after oral dosing to numerical comparison concentrations reported in neuronal and purified-protein experiments.

The analysis does not estimate a human brain concentration or a probability of toxicity. It shows which combinations of cross-system scaling, post-48-h retention, and dose history would be required for a numerical match, and which additional measurements would be needed for biological interpretation.

• src/analyze_annonacin_pre_qivive.R: complete analysis and figure-generation script.
• data/curated/: the two small input tables used by the script.
• data/search/: archived PubMed ESearch JSON responses, screened PMIDs, the title/abstract-level decision ledger screening_decisions.csv, and a script for retrieving current PubMed records.
• source_manifest_annonacin.csv: source locations and the role of each retained input.
• .Rprofile, renv/, and renv.lock: project-local activation and exact R package versions used for the archived release.
• outputs/: exact primary and supplementary result tables, input checksums, and R session information.
• figures/: the submitted main figure in TIFF and a PNG preview.
• SHA256SUMS: release-wide SHA-256 checksums, regenerated after the analysis outputs are finalized.
• THIRD_PARTY_NOTICE.md and LICENSE_RENV: provenance and reuse terms for archived bibliographic records and the bundled renv activation script.

The versioned .Rprofile activates the project-local renv library. From the repository root, restore the recorded environment once:

Rscript -e 'renv::restore(prompt = FALSE)'

Then run:

Rscript src/analyze_annonacin_pre_qivive.R

The script uses paths relative to the repository root and creates the outputs/ and figures/ directories if needed. It uses no random sampling. The lockfile records the tested package versions; the full runtime report is in outputs/R_session_info.txt. Raster generation uses ragg explicitly. Byte-identical figure reproduction also requires the recorded R version, compatible system libraries for ragg, and an installed Helvetica font; numerical tables do not depend on font rasterization.

After finalizing all files, regenerate and verify the release manifest:

python3 src/write_sha256_manifest.py
python3 src/write_sha256_manifest.py --check

The primary oral input is a published group mean of 1.4 ng/g in rat whole brain 48 h after a single 10-mg/kg dose. Error bars were not digitized. The targeted search identified no published quantitative human measurements of brain entry, regional or subcellular partitioning, or post-48-h retention. The displayed value M_B=1 is therefore an illustrative reference, not an estimate.

Please cite the versioned release using the metadata in CITATION.cff:

Dibot N. 2026. What Sparse Brain Kinetic Data Can—and Cannot—Tell Us About Oral Annonacin: Reproducibility Package. GitHub. Version 1.0.0. https://github.com/NicolasDibot/oral-annonacin-pre-qivive/releases/tag/v1.0.0.

Once the accompanying article is published, please cite the article as well.

The repository LICENSE applies to the original code, documentation, derived tabular outputs, and original figures in this release. It does not apply to PubMed/NLM metadata, source publications, or the third-party renv activation script. Their provenance and reuse terms are described in THIRD_PARTY_NOTICE.md; the renv license notice is reproduced in LICENSE_RENV. Values extracted from published sources remain subject to their original terms, and the manifest links to each source. No publisher abstract or third-party full-text article is redistributed in this release.