Skip to content
gabeykimPublic

About

CROSSHOST: a benchmark for predicting bacterial regulatory DNA activity across host species

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

45 Commits

Folders and files

Repository files navigation

CROSSHOST

A benchmark for predicting bacterial regulatory DNA activity across host species. See package/README.md for the full project description, central finding, licensing quarantine, and quickstart — this file exists only to get a fresh clone of the repository (not the curated package/ distribution) working, since nothing at the repository root pointed to that file before Gate 14 found the gap.

Repository layout, in brief: scripts/ and data/ are where the actual pipeline lives and runs; package/ is a curated export of selected outputs, built by make package, not a self-contained pipeline in its own right (see the Makefile's own header comment). Most people cloning this repo want to run make audit / make reproduce, both of which run from the repository root, not from inside package/.

Setup

python3 -m venv .venv
source .venv/bin/activate
pip install -e ./package                # NOT `pip install -e .` -- there is no pyproject.toml at the repo root
pip install -e "./package[citations]"    # + certifi, needed only for `make verify-citations`

requires-python = ">=3.10" (declared in package/pyproject.toml). Verified working on Python 3.14.2, macOS/Apple Silicon, as of Gate 14 (out/results/gate14_clean_clone_verification.md) — no other Python version or OS has been tested.

Checks

make audit             # leakage + provenance audits -- fast, offline
make verify-citations   # mechanically verify every manuscript citation -- live network (CrossRef/arXiv)
make test               # package test suite
make reproduce          # regenerate all figures/tables from shipped intermediate results -- fast
make reproduce-full      # full pipeline from raw/ -- MANY HOURS, not re-tested end-to-end, see Makefile header

Run make help for the same list with one-line descriptions.

Publisher material not redistributed

Five Johns et al. (2018) files are not included in this repository, because they are Springer Nature / Nature Methods copyright and this project does not hold redistribution rights. Availability on PMC under the NIH Public Access Policy is not a redistribution licence.

file size what it is
raw/nihms945382.pdf 1.2 MB the full author manuscript
raw/nihms945382.nxml 108 KB its full text in XML
raw/NIHMS945382-supplement-3.xlsx 8.4 MB supplementary table
raw/NIHMS945382-supplement-4.xlsx 5.4 MB supplementary table
raw/NIHMS945382-supplement-5.xlsx 7.6 MB supplementary table

Download them from the publisher and place them in raw/ under exactly those filenames:

You only need the three supplements for make reproduce-full. make audit, make test, make verify-citations and make reproduce (the fast path) all run from shipped intermediate results and read none of these files. The scripts that do are scripts/01_inspect_supplements.py, scripts/02_build_core_tables.py (supplements 3 and 4) and scripts/72_attenuation_analysis.py (supplement 5). No script reads the pdf or the nxml — they were present for reference only.

Kept for provenance-audit purposes: nihms945382_bodytext.txt (the extracted body text, used as a unit-test fixture by scripts/03 and scripts/15), the five main-text figure thumbnails, and the two smallest supplements (-6, 46 KB and -7, 14 KB).

Reproducibility

Verified from a genuine fresh git clone (not an existing working directory, not an isolated venv reusing a checkout already on disk) as of Gate 14: make audit, make verify-citations, and make reproduce all pass with no undocumented step beyond what is written in this file. Full verification record, including every gap found and fixed to get there: out/results/gate14_clean_clone_verification.md. make reproduce-full was not attempted (would cost the ~40 cumulative hours of training this project's Gates 4–8 already spent) — see out/PREPRINT/REPRODUCIBILITY.md for what that does and does not mean.

The manuscript

out/MANUSCRIPT.md (working copy) and out/PREPRINT/MANUSCRIPT.md (submission-formatted, with figure numbering) are the same content. Start there for the actual scientific writeup; this README is setup instructions only.

About

CROSSHOST: a benchmark for predicting bacterial regulatory DNA activity across host species

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages