Add ncountr

[princello]

Add ncountr to the scverse ecosystem

ncountr is a Python package for end-to-end NanoString nCounter gene expression analysis. It provides RCC file parsing, quality control, normalization (positive-control, CodeSet content, and housekeeping), differential expression testing (per-gene Mann-Whitney U or t-test with FDR correction), gene set scoring, and GSEA.

AnnData integration

nCounter is a bulk, targeted-panel platform, so the parsing, platform-specific QC and normalization stages operate on ncountr's own NanostringExperiment container. AnnData is the interoperability layer: the module-level function ncountr.to_anndata(exp) exports the analysed experiment — normalized counts in .X, raw counts in .layers["raw"], sample metadata plus QC metrics and lane/cartridge information in .obs, and gene annotations including housekeeping flags in .var — so the result can be handed to scanpy and the rest of the scverse ecosystem without losing its sample or gene metadata.

The GSE275334 Long COVID vignette demonstrates that handoff end to end: PCA, principal components coloured by the QC and lane variables carried across in .obs, a group dendrogram, rank_genes_groups, and a matrixplot — plus a cross-check that re-running the same contrast in scanpy reproduces the ncountr.de() gene ranking (Spearman ρ = 0.999998, 19 of the top 20 genes shared).

Checklist

• OSI-approved license: MIT
• Versioned releases: published on PyPI
• CI/CD: GitHub Actions (workflows)
• API documentation: hosted on Read the Docs
• AnnData integration: ncountr.to_anndata(exp) export

The full checklist from the repository README is filled in in this comment.

[muskanhashim]

Thanks for the submission! I reviewed the code against our checklist and it's in great shape

We would like to flag, though, that the PR uses a shortened checklist, and it's missing the mandatory item

"I am an author or maintainer of the tool and agree on listing the package on the scverse website"

Could you paste the full checklist from the repo README and tick all mandatory items including this? That author-consent confirmation is required for us to list the package

Two more minor, non-blocking notes: the "automated tests" checklist item is also missing, though your tests clearly cover it; and the version field reads "v0.1.0" while PyPI lists it as "0.1.0". You may want to drop the "v"?

Other feedback (not related to the checklist): currently AnnData only comes in at the end via to_anndata(), while the analysis runs on the NanostringExperiment object. That's completely fine but it would just be nice if a tutorial showed how to take the exported AnnData onward into a scanpy workflow, so users can see how ncountr plugs into the broader scverse ecosystem!

Thanks again!

[princello]

Thanks for the careful review, @muskanhashim — apologies for the shortened checklist. Here is the full one from the README with all mandatory items answered.

Also fixed in packages/ncountr/meta.yaml: the version field now reads 0.2.0 — I dropped the v as suggested, and bumped it to match the current PyPI release (0.2.0 was published after this PR was opened).

One correction to my own PR description above: I wrote NcountrExperiment.to_anndata(), which is wrong on both counts. The container class is NanostringExperiment, and to_anndata is a module-level function, so the actual call is ncountr.to_anndata(exp).

---

Checklist for adding packages

Mandatory

Name of the tool: ncountr

Short description: A Python package for end-to-end NanoString nCounter gene expression analysis — RCC file parsing, quality control, normalization (positive-control, CodeSet content, and housekeeping), differential expression testing (per-gene Mann-Whitney U or t-test with FDR correction), gene set scoring, and GSEA.

How does the package use scverse data structures: ncountr parses raw RCC files into a NanostringExperiment container, and exports the complete analysis state to AnnData via ncountr.to_anndata(exp): normalized counts in .X, raw counts in .layers["raw"], sample metadata plus QC metrics and lane/cartridge information in .obs, and gene annotations including housekeeping flags in .var. AnnData is the handoff point into scanpy and the rest of the scverse ecosystem for downstream analysis. NanoString nCounter is a bulk, targeted-panel platform (typically 50–800 genes), so the raw-parsing and platform-specific QC/normalization stages operate on the native container, and AnnData is used where it is the appropriate structure — as the interoperable representation of the analyzed experiment.

• The code is publicly available under an OSI-approved license — MIT (LICENSE)
• The package provides versioned releases — 0.1.0 and 0.2.0
• The package can be installed from a standard registry (e.g. PyPI, conda-forge, bioconda) — PyPI: pip install ncountr (a conda-forge recipe is also in review at Add ncountr conda-forge/staged-recipes#32721)
• Automated tests cover essential functions of the package and a reasonable range of inputs and conditions — 313 tests across 18 test modules in tests/, covering every public module (parsing, QC, normalization, DE, effect sizes, GSEA, pathway scoring, gene sets, plotting, AnnData export, GEO download, CLI)
• Continuous integration (CI) automatically executes these tests on each push or pull request — GitHub Actions, on every push and PR to main, across Python 3.9–3.12
• The package provides API documentation via a website or README — full API reference on Read the Docs
• The package uses scverse datastructures where appropriate (i.e. AnnData, MuData or SpatialData and their modality-specific extensions) — AnnData, via ncountr.to_anndata(); see the description above
• I am an author or maintainer of the tool and agree on listing the package on the scverse website

Recommended

• Please announce this package on scverse communication channels (zulip, discourse, twitter)

• Please tag the author(s) these announcements. Handles (e.g. @scverse_team) to include are:

  • GitHub: @princello (Zicheng Wang, Columbia University Irving Medical Center)
  • Zulip / Discourse / Mastodon / Bluesky / Twitter: not currently set up — GitHub is the best handle to use.

• The package provides tutorials (or "vignettes") that help getting users started quickly — five vignettes on real public GEO datasets spanning 1,458 samples and five different nCounter panels (Long COVID / ME-CFS, HCC immunotherapy response, autoimmune retinopathy, interferon-stimulated macrophages, and a 1,321-sample PAM50 breast cancer set)

• The package uses the scverse cookiecutter template — not used; the project structure predates the submission, though it follows the same src/ layout, pytest, and Sphinx/RTD conventions.

---

On the scanpy tutorial suggestion — that is a fair point, and you were right that it stopped at the export. This has since been implemented; see the follow-up comment below.

Thanks again for the review.

(Edited: corrected the differential-expression description — ncountr.de() uses a per-gene Mann-Whitney U or t-test with FDR correction, not a negative-binomial GLM — and refreshed the test count.)

[muskanhashim]

Great, @princello! Everything from the checklist side is resolved, thanks for the thorough follow-up and the Scanpy vignette!

In our attempt for users to be able to discover scverse ecosystem packages more easily, there is unfortunately a small new requirement on our side: The registry schema was just tightened (#392) after your PR was opened, so the metadata validation (mkjson + pre-commit) now fails. Two quick fixes needed in packages/ncountr/meta.yaml to match the new schema:

1. tags must now come from the controlled vocabulary (the free-form ones no longer validate). For ncountr I'd recommend something like:
   tags:
   - bulk RNA-seq
   - quality control
   - differential expression
   - functional analysis
2. two new required fields:
   primary_category: bulk RNA-seq
   language: Python

Please feel free to pick whichever tags/primary_category best fit from the allowed lists in the schema. Once that's in and the checks go green, this is good to merge. Thanks again for bearing with the moving target!