HCS-proc v2

The maintained pipeline, corrected and extended since the article was published.

v1 remains the exact code used for the article. Two of the changes below alter
results, so cite v1
(10.5281/zenodo.17949848) to reproduce the
published run, and the concept DOI
10.5281/zenodo.17949847 to cite the pipeline
itself.

Method changes

These change results, deliberately, and are the reason v1 is kept frozen.

• The Brain-Cousens BC4 model is corrected in
  scripts/feature_selection/plots_emd_model_drc.py. BC4 is BC5 with the lower asymptote
  fixed at zero, and the f * x term is what makes it a hormesis model rather than a
  log-logistic. The published run's BC4 was not that curve, so AIC and BIC for BC4 differ
  from the article's.
• The highest exposure concentration is excluded from curve fitting, via the new
  drc_excluded_concentrations key. The EMD scores still cover every pair; only the
  fitting skips the excluded ones, so that concentration-response detection stays within
  sub-cytotoxic exposure. The script prints the pairs it is fitting when it starts.

New pipeline step

• Feature categorization, scripts/categorization/: build_feature_categories.py,
  count_feature_types.py and plot_types.py, with the feature_categories.tsv lookup
  and a [categorization] configuration section. It reports features it cannot resolve
  rather than dropping them silently, and counts lysosomal puncta as counts. The README
  gains a section and the figure it produces.

Configuration and documentation

• The violin plot script is configuration-driven, with a [violin_plots] section,
  matching the rest of the repository instead of the "modify file X" style.
• The README embeds the published figures (Figure 2 and Figure 4 of the article).
• Citation is complete: the article with volume, issue and page range, PMID and PMCID;
  CITATION.cff with the full author list, ORCIDs and affiliations; the code archive and
  the dataset, both by concept DOI.
• CITATION.cff now carries the pipeline's own version, release date, abstract and
  DOI, so this release's Zenodo record is described rather than auto-generated.
• The concentration-response feature selection has been extracted into a standalone tool,
  NIB-SI/cdr_FS
  (10.5281/zenodo.22047439), which the feature
  selection section now points at and cites. The scripts here remain the record of what
  was run for the article; use cdr_FS to apply the method to another dataset.

Data

Unchanged: https://doi.org/10.5281/zenodo.17951792 (CC-BY-4.0, ~121 GB), the concept DOI
the article's Data Availability statement cites.

HCS-proc

Release for article titled "A high-content imaging pipeline to investigate subcytotoxic effects in RTgill-W1 cells".